JPH05259127A - Method for etching monitoring in plasma etching - Google Patents

Abstract

PURPOSE: To obtain a technique for measuring a thickness of a thin film on realtime basis using a standard end detector by a method, wherein plasma radiation lights strength in selected wavelength is monitored, and a change in the plasma radiation light intensity is related to a remaining thin-film thickness. CONSTITUTION: This is a method for monitoring removal of a thin film on a wafer product during plasma etching, and plasma radiation lights strength in selected wavelength is monitored, and a change in the plasma radiation light intensity is related to a remaining thin-film thickness. For example, when an oxide on silicon is etched, plasma radiation lights intensity 302 as a function of an etching time is measured through a filter of 436±5 nm contains a periodic change caused by interference of lights. Meanwhile, a relationship between the etching time and a remaining thin-film thickness 404 is linear, and in the case of this example, an etching end occurs when 40 sec, passes, and at that time, dropping of radiation light strength is observed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the manufacture of semiconductor devices, and more particularly to plasma etching. Plasma deposition is also disclosed as an application of the invention.

[0002]

2. Description of the Related Art An important process in the manufacture of integrated circuits is the process of removing layers of various materials formed on a silicon wafer. As the removal process, generally, the following two types of important techniques are used. 1) Wet or chemical etching in which a patterned silicon wafer of photoresist is immersed in a chemical solution. 2) Dry or plasma etching in which the wafer is exposed to a plasma containing a gas such as CF ₄ . The present invention is mainly concerned with the latter plasma etching.

Plasma etching processes and devices are generally well known for materials that are etched in the manufacture of semiconductor devices. The process begins by applying a mask material such as photoresist on a silicon wafer. The mask pattern is for preventing a specific area on the wafer from being etched. The wafer is then placed in a plasma reactor (etching device) and etching is performed.
Subsequent processes are determined by the type of semiconductor device to be manufactured. Such a plasma etching process is effective for precisely etching a particularly small shape.

A typical etching process includes removing polysilicon (poly Si; hereinafter also referred to as "poly") coated on a gate oxide (hereinafter referred to as "oxide"). Of the various problems, it is important that the etching selectivity (poly: oxide selectivity) between polysilicon and the underlying gate oxide is as high as possible and that oxide loss is low. is there.

LAM4 with upper side power supply and wafer grounding
A typical etcher, such as a 90 etcher, achieves a poly: oxide selectivity of 100: 1 or greater using a low power (100 watt) Cl ₂ / He plasma. In this case, a severe undercut, that is, a notch is generated from the side wall of the polysilicon. On the other hand, according to the helium rich (more than 50%) and high power (more than 200 watts) process,
Vertical sidewalls can be obtained at the expense of selectivity (less than 30%), but with the associated tendency to damage the gate oxide. Side etch occurs during overetch, not during polysilicon removal. Therefore, the amount of overetch should be limited.

Regardless of the type of etching applied, detecting when the coating material ("film", "thin film") has been completely removed ("clear") is of wide and great interest. This is typically accomplished by detecting the presence of the underlying material exposed in the plasma using means such as spectroscopy.

It is also advantageous to know (being able to detect) when the material has been removed by the required amount-in some cases less than all of the coating material-.

US Pat. No. 4,415,402 discloses a method for detecting the complete removal of doped silicon dioxide from above an underlying substrate by plasma etching. There is.
This US patent shows a technique for detecting the end point by utilizing the fact that the intensity of a specific wavelength or a specific wavelength region changes when the material to be etched is completely removed. However, this U.S. patent does not disclose a technique for detecting when the thin film to be etched has a predetermined residual thickness due to etching.

US Pat. No. 4,454,001 discloses optical interferometry for measuring etch rate. This method uses a phenomenon in which a pattern etched in a substrate produces a diffraction pattern by incident light. In this method, a beam of light is incident on the etched area of the substrate. The light reflected from that area forms a diffraction pattern whose diffraction intensity is detected and recorded as a function of time during the etching process. The substrate etch rate is inversely proportional to the period of oscillation in the recorded intensity / time curve.

US Pat. No. 4,680,084 discloses optical interferometry for monitoring etching and measuring thickness. In this method, light is projected onto the area of the substrate where etching is in progress.
By detecting the intensity of the reflected light, the etching depth of the area where the etching is in progress is monitored,
Alternatively, the thickness of the area is measured.

US Pat. No. 4,687,539 discloses endpoint detection and control in a laser etching process. In this method, an excimer laser vaporizes a continuous layer of chromium over an area, and chromium chloride reaction products are produced on the area to be etched. A die laser is projected onto a zone above the area to be etched and the copper chloride in that zone fluoresces. The narrow band photodetector detects the fluorescence.

US Pat. No. 4,717,446 discloses end point detection in etching epitaxially grown silicon. In this way, "in process"
The "monitor" wafer is etched side by side with the above wafer. The monitor wafer consists of a silicon substrate having an oxide layer and a polysilicon layer coated thereon. The laser is used to measure the etch rate of the monitor wafer by measuring the reflected light at the oxide layer.

US Pat. No. 4,936,937 discloses a method for endpoint detection in plasma processing.

In US Pat. No. 5,023,188, an optical interferometry method is disclosed, in which the etching groove depth is detected by irradiating a wafer surface with coherent light.

US Pat. Nos. 4,846,928, 4,
Nos. 847,792 and 4,861,419 commonly disclose a method for comparing endpoint traces obtained by conventional means with reference traces to detect anomalies. .. The reference is used to determine the area of the endpoint curve (ie, the curve before the endpoint and the endpoint, over the duration of the overetch) and to know the slope of the curve within that area.

From the above-mentioned conventional techniques, it can be generally said as follows. That is; 1) various techniques and devices are known for performing monitoring of the plasma etching process; 2) another irradiation light (eg a laser) is used for monitoring; 3) etching characterization and / or Alternatively, a test (monitor) wafer may be used for monitoring; 4) Endpoint detection is basically performed by monitoring removal of a thin film (layer) being etched.

Considering the etching process itself, various etching "recipes" are used. For example, the flow rate, pressure, electrode spacing, gas type can be varied depending on the material being etched to provide the required selectivity, etching rate, and uniformity.

[0018]

SUMMARY OF THE INVENTION Accordingly, one object of the present invention is to provide a technique for measuring thickness in real time using a standard end point detection device.

Another object of the present invention is to provide a technique for observing an optical interference effect in plasma etching by using a passive monitoring device without using an active irradiation source (for example, another laser beam). To provide.

Still another object of the present invention is to obtain a higher selection ratio from an etching process having a relatively high etching rate and a low selection ratio when the thin film is etched to a certain residual thickness during the etching process. It is to provide a technique for switching to an etching process.

Yet another object of the present invention is to provide a technique for determining etch rate uniformity, average etch rate, and selectivity for a wafer being processed (ie, without a monitor wafer). It is in.

Yet another object of the present invention is to provide a technique for characterizing etching parameters that are self-scaled.

Still another object of the present invention is to provide a technique for constructing an in-line type etching apparatus without using a test (monitor) wafer, that is, an offline type thickness measurement. It is in.

[0024]

SUMMARY OF THE INVENTION To solve the above problems, the present invention is basically a method for monitoring thin film removal on a wafer product during plasma etching: Monitoring plasma radiant light intensity at a wavelength; correlating changes in plasma radiant light intensity with residual thin film thickness.

[0025]

In accordance with the present invention, the intensity of the emitted light from the plasma at a particular wavelength is monitored during etching and the plasma emitted light intensity is related to the residual thickness of the thin film being etched. With this method, it is possible to detect when the thickness of the thin film remaining on the substrate reaches a certain thickness.

According to one aspect of the invention, the overall etching process can be scaled during etching by knowing when the residual thin film thickness reaches a certain thickness. In this way, by knowing the etching rate and the residual thickness, the end point of etching can be predicted with high accuracy, and over-etching (over-etching) can be carefully controlled.

In another aspect of the present invention, the thin film is etched by an etching method with a high etching rate and a low selection ratio at an initial stage, and when it is detected that a certain residual thickness is reached, the etching method is performed. A low etching rate, high selectivity etch technique can be switched to more carefully control overetching, undercutting and damage to the underlying substrate.

In the present invention, by using the plasma itself as the light source and as the means for transmitting the thickness information, it is possible to avoid the use of another additional active irradiation source. This is extremely useful if the etching has to be stopped before the underlying material is removed. This technique also monitors the plasma deposition process (ie not plasma etching).
Is also useful.

Further, according to the present invention, the in-line process for determining the etching rate, the uniformity and the selection ratio can be facilitated by monitoring the intensity of the plasma emitted light.

Other objects, aspects, and advantages of the present invention will be apparent from the following description.

[0031]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention relates to monitoring of plasma etching, and the apparatus itself is well known. Therefore, the following description will be made mainly on the monitoring technology itself rather than individual apparatuses. To do.

[Optical Interference in Monitoring End Point of Plasma Etching] In the conventional method using the reflection spectroscopy (optical interferometry), the thickness of the thin film or the thin film laminated body is determined by comparing the thickness of the thin film or the thin film laminate with that of the reference substrate. It is determined by measuring the relative amount of reflection. The reference substrate is usually silicon or aluminum,
A material that is totally reflected is used. The multiple reflected lights with respect to the incident light are combined to generate a reflectance that depends on the thin film thickness. The interference pattern also changes with wavelength. Various thin film thickness gauges use this effect.

FIG. 1 illustrates the mechanism of interference resulting from multiple reflections within a transparent thin film on a reflective substrate. Incident light 102 is refracted in the thin film 104 and reflected on the surface of the lower substrate 106.
Ray 108 represents the light emitted from the thin film after being reflected once by the substrate. A part of the light reflected once is reflected inward by the upper surface of the thin film and returns to the substrate side. Ray 110 represents the light emitted from the thin film after being reflected twice by the reflective substrate.

FIG. 2 shows the measurement results of the oxide thin film on the silicon substrate when the thickness of the oxide thin film is reduced by etching, and the relative ratio (%) of the reflection amount of the oxide thin film (vertical axis). ) Represents the change as a function of the oxide thin film thickness (residual thickness). The measurement is Wild Leit
z) The measurement was carried out at a wavelength of 436 nanometers (nm) using an LTS-M / SP measuring device. In the data plot,
The 100% reflection corresponds to the reflection of an uncoated theoretical silicon wafer (reference silicon). The periodic change in the data curve shown in FIG.
It is due to.

FIG. 3 is for etching oxide on silicon (see FIG. 2), where the actual plasma radiant light intensity (squares) 302 and the theoretical% reflectance (as a function of etching time) ( Black dots) 304 are shown. Plasma radiant light intensity is indicated on the vertical axis by "count value"
As shown. The data shown in this FIG.
w) Obtained while etching blanket plasma LTO (low temperature oxide coating) on a 4500 etcher. A 436 ± 5 nm filter was placed behind the window of the back chamber.

As shown in FIG. 3, the trace of the plasma emission light intensity (trajectory: square mark 302) is rapidly settled by a count value of about 400 when the etching time is 24 seconds. The oxide is initially 2500 Angstroms thick and uniform (less than ± 2%), and this pattern is reproducible from wafer to wafer. The actual etching end point occurs after 40 seconds, at which time the oxide is removed and the lower silicon appears, and the count value drops by 200.

The shape of such a pattern is similar to the pattern due to optical interference shown in FIG. 2, which is due to the fact that the oxide of the wafer absorbs a considerable amount of light from the plasma. it is conceivable that.

Returning to FIG. 2 again, the maximum absorption occurs at the time when "204" is added at the oxide thin film thickness of 800 Å. The thickness scale (horizontal axis in FIG. 2) is converted into the etching time scale (horizontal axis in FIG. 3) as the oxide thickness that changes with etching.

FIG. 4 shows the relationship between the etching time and the residual thin film thickness. As shown in FIG. 4, the etching rate is linear with time. Using this function and the data in FIG. 2, the theoretical relative reflectance% can be calculated as shown in FIG.
It can be plotted as (black dot 304) overlaid on the plasma emitted light intensity curve (302). Although the absolute values themselves are not comparable, the relative positions of the maximum absorption points are well matched (ie, 24 seconds for the plasma emission intensity curve 302 and 27 seconds for the theoretical curve 304; in FIG. 4). See point 404). The deviations from each other may be due to lack of assumptions about reflectance, etch rate changes, and measurement error.

The interference effect is due to the use of a very uniform oxide over a large planar area and applying a uniform etch rate, as is the case for blanket oxide on silicon as used herein. , Will appear more prominently. On the other hand, the degree of interference effects is reduced when etching patterned non-uniform oxides with comparable average thickness (50 in FIG. 5).
4) has been confirmed.

FIG. 5 shows a uniform plasma radiant light intensity (black dots 504) from a blanket oxide and a patterned non-uniformity when etching is performed by a typical oxide etching method using a rainbow etching apparatus. The radiant intensity of plasma emitted from various oxides (diamond-shaped black dots 502) is shown at a wavelength of 436 nm.

These observations were unexpected because the apparatus for collecting plasma radiation could not see directly above the wafer and was somewhat obscured by the focus ring surrounding the plasma contour. ..
The crystal window allows light to pass through the ring.
This means that the oxide absorbs some amount of the emitted light from the plasma at that wavelength.

As already mentioned, the behavior of the end-point traces can be observed during the etching of the oxide, by simply monitoring the emitted light intensity from the plasma and the wafer, for example at 436 nm ± 5 nm. This behavior is obtained by using a reflectance spectrophotometer (for example, the basis of operation of the Wildlights device for measuring thin film thickness). The intensity of emitted light due to the optical interference effect correlates well with the theoretical reflection amount value obtained by the Wild Lights thin film measuring device. This effect stops the etch (end point) before the coating on the underlying material disappears,
It can be used to switch to an etching process (etching method) having a higher selection ratio and less damage to the substrate.

Very generally speaking, real-time film thickness measurement during plasma etching is not new per se, but has hitherto required complicated optics, and usually additional additional irradiation sources. (For example, laser light) was required. According to the invention, by monitoring the emitted light intensity of the plasma-wafer combination itself,
Useful information about the etching process can be obtained without the need for separate equipment.

Thus, according to the present invention, the end point trace reveals the optical interference effect caused by the plasma radiation absorbed by the thin film being etched. The minimum point of absorption corresponds to the detection of a certain residual thickness, which should therefore be used to stop the etching before the coating film on the underlying material disappears (ie before it is "removed"). You can From this,
It becomes possible to switch the etching method to a process having a higher selection ratio and less damage at a predetermined thickness before being completely removed. Also, the method works better with thin films and regions with very uniform etch rates and large areas (unpatterned wafers).

This "passive" method requires no special hardware other than the standard endpoint devices (filters, monochromators, diode arrays) found in our plasma etching systems. In each application, it will be necessary to carefully and carefully select the wavelength. Wavelengths corresponding to strong emission from excited gas products should be avoided as they mask the interference effects.

The effect is observed while etching the oxide on silicon. Other applications are currently under study. For example, it can be applied to the measurement of thickness during plasma deposition.

Results such as those already mentioned have been obtained and demonstrated during the etching of oxides on silicon, but those skilled in the art closest to this invention will be familiar with plasma etching of other materials. It can be appreciated that similar results should be obtained for control. Furthermore, although the above discussion basically relates to plasma etching, it can be appreciated that similar results can be obtained for the plasma deposition process.

[Application of Synchrotron Radiation Spectroscopy] [Characteristics of Etching Process] The dry etching process is systematically characterized in order to evaluate whether or not the work conforms to a predetermined specification (work range). Have to be done. If the work being performed deviates from the predetermined work range, the device must be adjusted or repaired.

Solving process-related problems and characterizing processes involves statistical techniques such as test design techniques (DOE), response surface techniques (RSM) or Taguchi techniques among various known techniques. Solved by using.

Whether or not the process is proceeding within specifications is typically determined by whether or not more inspection items are produced in less time, which is a "process suitability assessment". Sometimes called. Qualification should be fast enough to reduce manufacturing delays, although fully sufficient to detect (warn) defects.

Suitability in etching is typically done by conventional techniques on test (non-machined, non-manufactured) wafers, etching rate, etch uniformity and, in some cases, selectivity or linearity. Evaluated for width measurements. Uniformity and selectivity are determined from etch rates measured at various locations across the wafer using an optical interferometer (diffraction spectrometer) or ellipsometer for thin film thickness measurements before and after etching. .. Since the test wafer must be partially etched, it is not desirable to re-etch the wafer in the actual manufacturing process until it is completed from the viewpoint of quality. Therefore, use it for the aptitude evaluation as described above. Is not preferable.

FIG. 6 and FIG. 7 show an outline of two types of aptitude evaluation systems. FIG. 6 shows a flow chart of the steps including off-line qualification of the etching process. FIG. 7 shows a flowchart of steps including real-time and in-line qualification of the etching process. The latter aims to save time and reduce test wafer costs.
It is controlled by the wafer level.

In addition to the off-line test wafer method described above, another technique for analyzing etch uniformity and endpoint tracing using optical radiation means. Exists. this is,
"Monte Carlo Simulation of Plasma Etching Radiation Endpoint" (E. J. Pawlek, Emerging Semiconductor Technology, ASTM S)
TP 960, Dee. C. Gupta and Pea. H. See Langer Editor, American Society for Testing and Materials, 1986).

Plasma radiation itself is traditionally used for etching endpoint purposes. For example, when the etching of the uppermost layer is completed, the lower material is exposed, and the radiation intensity at a certain wavelength changes. Such a single signal change is used as a means to stop the etch and protect the underlying material. This is US Pat.
No. 312,732.

According to the technique of the present invention, it is possible to eliminate the need for off-line measurement, which requires stopping the actual manufacturing work, or to reduce the off-line aptitude evaluation time for obtaining necessary data. Therefore, the suitability evaluation time and cost can be reduced.

As a result, a method was developed based on analyzing the optical emission signal from the plasma during etching. Additional thickness measurements are still needed, but they are part of a typical manufacturing process (eg, pre-etch film thickness, post-etch bottom film thickness, etc.). ..

In general, in view of the non-uniformity of the thin film and the non-uniformity of the etching, the etching is carried out beyond the end point (excessive etching) to ensure complete removal.
To be done. If a long overetch is needed,
In order to protect the underlying material (eg, to protect the thin gate oxide), the etch selectivity ratio for the top thin film material: bottom material must be high.

Therefore, in order to reduce excessive etching, it is important to have a high selection ratio in order to protect the material on the lower side, and at the same time, to improve the uniformity of the thin film and the uniformity of etching. The etching rate itself affects the processing efficiency.

The above-mentioned “Monte Carlo simulation”
As described in the prior art regarding the Monte Carlo method, the thin film of the wafer was cleared (end point).
It is used to simulate the intensity of emitted light at a time point. Although some assumptions have been made, one of them (No. 4) is significantly different from the present invention in the following points. That is, in the case of the prior art, the product species (pro
exposed (lower side)
While it is assumed to be linearly proportional to the film area, the present invention states that the concentration of the product species is not only proportional to the exposed area but also relates to the etching rate. Is. In addition, the prior art assumes a random Gaussian dispersion for the instantaneous etch rate, whereas the etch rate is generally not reproducible for position within the wafer (eg, center to edge). It is a function. After all, in the case of the prior art,
The aim is to reproduce the radiant light intensity at the end point under a certain etching condition. On the other hand, in the present invention, it should be emphasized that the etching situation is derived from the analysis of the synchrotron radiation curve under certain conditions.

The symbols used in the following description and drawings have the following meanings. I Plasma radiant light intensity S Monitored signal t Time [X '] Concentration of species X in excited state [X] Concentration of species X in non-excited state A All exposed regions without etching mask a Thin film removed region E will be subject to overetching E Average etching rate averaged over the wafer when the thin film is etched E ′ Average etching rate averaged over the wafer when the underlying material is etched sigma _y species concentration of the total amount k _y species Y about second-order rate [Y] type Y of Y k _e electron impact dissociation rate k _p feed rate c, b, k, β, α, B constant S ₀ monitored before the end point signal strength S _∞ completely time H initial thin film etching is completed monitor signal intensity t ₁ film removal start time t ₂ wafers of the wafer after removal of the average film thickness U _n initial film uniformity Uniformity E _O radius Δt minute time increment Δr small radius increment δ variables T ending position R wafer etch rate r radial direction in the end portion of the etch rate E _R wafer at the center of the wafer (radial direction) of the etch rate E Time until reaching T _E Total etching time including excessive etching d ₀ Initial lower layer thickness d _m Minimum thickness of lower layer after etching d _a Average thickness of lower layer after etching R _m Etching Minimum selection ratio between rates E and E'R Average selection ratio between etching rates E and E'i ₀ , i _e , S _e , s ₀

According to the invention, a filter, monochromator, or diode array is used to collect the plasma radiation at a given wavelength or wavelength range. The emitted light intensity I is converted into an electric signal (voltage), digitized (counted), and the final signal S is monitored.

In the technique of the present invention, the following ten assumptions (numbers 1 to 10) are established.

Assumption # 1: The invention is best applied to single wafer etchers, but is also valid for batch systems. The most recent dry etching equipment is
For single wafer process.

Assumption # 2: It is assumed that the emitted light intensity I represents not the specific part of the whole plasma but the whole plasma. In a single wafer system, the volume of the plasma typically forces its diffusion to the concentration of species monitored over the timescale of sampling the emitted light (eg, sampling rate exceeds 30 msec). Small enough to average out to.

Assumption # 3: The radiation signal S is assumed to be proportional to the radiation intensity I. That is, if t is time,
S (t) §I (t) (where “§” represents a proportional symbol; the same applies hereinafter). The detector has sufficient linearity (<±
5%) and have an adequate effective range to avoid saturation.

Assumption # 4: The monitored signal should preferably be representative of radiation from species produced by etching (eg oxygen in an oxide etch) or consumed (eg fluorine in a silicon etch). .. Assuming that [X] represents the concentration of the species in the gas phase and [X '] represents the concentration of the species in the excited state, it is assumed that I (t), [X'], and [X] are all proportional. To be done. That is, I (t) § [X ′]
§ [X]. Such an assumption will be realized if the electronic excitation cross section of the monitored species is stable. If the electron concentration at the excitation energy of the species being monitored is not constant, a photometer can be used to compensate for changes in the electron energy distribution. Photometers are described in "Optical Emission Spectroscopy in Reactive Plasmas: A Method for Correlating Radiant Intensity with Reactive Particle Density" (J. W. Cobain, M. Chen, J. Appl. Phys., 51,
pp3134-3136, June 1980).

Assumption # 5: Losses on excited species are negligible except by relaxation. That is, the supply rate constant is much smaller than the excitation rate.
Under the above assumptions, the monitored species concentration [X] can be written as:

[Equation 1]

Here, "[X]" is a value averaged over the entire volume of plasma, "c" and "b" are constants, "A" is the total exposed area, and "a". Is the area removed, "E" is the average film etch rate across the wafer, and "E" is the average etch rate of the underlying material. And k = Σ _y k
_y [Y] -k _e k _p is the production-loss balance of the species X due to the secondary gas phase reaction with the concentration [Y] for the species Y and the velocity k _y , the rate of electron collision dissociation. with k _e and feed rate k _p . (Note that k _e occurs for molecular species only).

Assumption # 6: k is assumed to be constant throughout the process. This assumption, thin removal of the wafer, is effective when X is not give a significant effect on the electron impact dissociation and concentration [Y] and rate k _y in a cross section in the case of a molecule.

When etching products are monitored, their concentration should be a function of the etching rates E, E '. If instead the consumable substance is monitored, the same expression can be made, but the concentration is −
It is proportional to [aE + (A−a) E ′].

In addition, there will be some radiation when the thin film on the wafer is removed, ie when a = A. Also, when there is no wafer, or when the wafer is patterned by a mask with a non-zero etch rate, there will be background radiation. Using Equation 1, the monitored signal "S" can be written as Equation 2 below.

[Equation 2]

Where α = a / (1-k) and β = b
(1-k) is a constant, and “B” is A = 0 and a =
This is the background signal for the case of 0. The parameter "B" also counts for gas phase reactions.
Before the end point, a = 0, and after completely etched, a = A. Defining S ₀ and S _∞ as S ₀ = αAE + B S _∞ = βAE + B and considering the derivative S (t) with respect to time, we have

[Equation 3] Can be obtained. If A and B are counted, the etching rate is constant. If there is a loading effect, the etch rate will be a function of the area removed. Further, the etching rate becomes a function when the etching conditions drift.

Assumption # 7: The etch rate is assumed to be constant if there is no significant loading effect. The loading effect can be reduced by increasing the power output and decreasing the residence time of the gas in the reactor (eg high gas flow rates).

Under the assumption of # 7, Equation 3 will represent the most common form of endpoint trace as a function of the removal pattern over time.

FIG. 8 is two graphs showing the non-linear removal pattern for the exposed area A, and FIG.
2A and 2B are two graphs showing a linear removal pattern for the exposed area A. If t ₁ and t ₂ represent the removal start time and the complete removal time by etching on the wafer, respectively, the following equation 4 can be obtained from the definition.

[Equation 4]

_Let "H" be the initial thickness of the film to be etched with uniformity U _n , and "E" be the average etch rate across the wafer when it is etched with uniformity "u". If defined as: t ₁ = H (1-U _n ) / (E (1 + u)) t ₂ = H (1 + U _n ) / (E (1-u)), the fastest etching rate and the thinnest It can be defined as representing the time required to remove a portion and the time required to remove the thickest portion with the slowest etching rate.

Assumption # 8: The analysis is limited if it can be assumed that da / dt = constant. That is, it is limited to the case where the end point trace changes linearly from S ₀ to S _∞ (see FIG. 9). This is generally not always true in the entire transition region from S ₀ to S _∞ ,
If the transition of almost the entire region can be approximated to a straight line, the practical purpose can be sufficiently satisfied.

FIG. 10 is four graphs simulating the end point curve in the case of performing thin film removal of a wafer at a radially symmetrical etching rate. FIG. 10 shows a simulation pattern of the end point trace (depending on whether the etching rate at the center is fast or slow) accompanying the removal from the center to the end. The etching rate E is
E = E _o ± r (E
It represented the _O -E _R) / _R. Here, “r” is the position in the radial direction (radial direction), “R” is the radius of the wafer, and “E” is
_O ″ and “E _R ” represent the etching rates at the center (O) and the edge (R), respectively, where a perfectly uniform thin film (ie u = 0) is assumed. Incorporating into Equation 2 above, and S ₀ = 1 and S
_∞ = 0 is standardized. The end point curve can be expressed by standardizing t ₁ to _t 2 as 0 to 1 in the same manner.

From this, the following equation can be derived from FIG. That is, in the case of the left side of FIG.

[Equation 5] In the case of the right side of FIG. 10,

[Equation 6] Becomes

These behaviors are 2πrΔ between the time t and t + Δt (however, t ₁ <t <t + Δt <t ₂ ).
This can be intuitively recognized by considering the case where the ring (r, r + Δr) in the region r is removed by etching. When the thin film removal of the wafer starts from the center and goes to the end, the drop toward the end becomes faster with the increase of the ring-shaped region (see the left side of FIG. 10). If the thin film removal pattern on the wafer is from the edge to the center, the opposite is true for the drop towards the end point (see right side of FIG. 10).

An endpoint curve similar to that of FIG. 10 results from loading effects and other removal patterns, which are not effective in determining etch rate uniformity.

Assumption # 8: da / dt = A (t ₂ −t ₁ ).
Can be written. Incorporating this relationship into Equation 3 using t ₁ and t ₂ defined as described above gives Equation 5 below.

[Equation 7]

Solve for u and δ = 2HdS / dt
The following Expression 6 can be obtained by using / [E (S _∞ −S ₀ )].

[Equation 8]

If H and U _n cannot be measured before etching, δ can be determined assuming that the time T to the end point is equal to E / H. Furthermore, T =
Since (t ₁ + t ₂ ) / 2 (see FIGS. 8 and 9), δ = 2TdS / dt / (S _∞ −S ₀ ) is eventually satisfied. Here, if the initial thin film is extremely uniform and δU _n << 1, the term of δU _n disappears from Equation 6, and in this case, the uniformity of the thin film needs to be measured. Disappear.

The etch rate selectivity between the film to be etched and the underlying material can be estimated if the underlying material thickness is known prior to etching and can be measured after etching. it can. Initial thickness is d
₀ and the minimum thickness after etching is d _m ,
When the selectivity is the worst, the selection ratio R _m is R _m = E (1-u) (T _E −t ₁ ) / (d ₀ −d _m ). Here, T _E is the total etching time including overetching. When defined in this way, the minimum selection ratio R _m becomes a conservative value for the average selection ratio R = E / E ′> R _m . The average selectivity selection ratio can also be estimated by _{R = E (T E -T)} / (d 0 -d a). Where d _a is the average residual thickness.

In order to derive the uniformity "u", it is supposed that the thickest portion of the thin film is etched at the slowest etching rate, and the thinnest portion of the thin film is tested at the highest etching rate. This is not always true, but generally H (1-U _n ) / (E (1 + u)) <t ₁ <H (1 + U
_{n) / (E (1 +} u)) and _{H (1-U n) /} (E (1-u)) <t 2 <H (1 + U
_n ) / (E (1-u)). Here, the sign (+ or −) before the “U _n ” term determines the region to which the transition applies. The sign before the "u" term can be positive or negative depending on the definition.

From Equation 6, the upper limit of the uniformity u is given.
The target value of the uniformity u is u = 0 when it is desired that the etching rate is completely uniform. However, this equation shows that when the uniformity of the thin film is poor, that is, δU _n ≧
If 1, it would not be possible to give significant data. In that case, the "u" term is estimated to be U _n (ie u → U _n ).

FIG. 11 shows a typical endpoint trace according to the present invention. The least-squares fit method is S ₀ and d
Used to determine S / dt. Time t ₁ is a straight line of S (t) = i _e + S _e t during the end point period and S before the end point.
(T) = i ₀ + S ₀ Obtained from the intersection with the straight line t, t
₁ = (i ₀ −i _e ) / (S _e −S ₀ ). Also, from the time when S departs from the gradient during the end point period, t ₂ = (S ∞ −
i _e ) / S _e is obtained. Further, S ₀ = i ₀ + S _o t can also be obtained.

In FIG. 11, the radiant light intensity is expressed in arbitrary units by time resolution of 0.1 second, and the end point (T
Indicates that there is a signal drop at approximately 62 seconds), and there is a slight overetching.

Assumption # 9: The monitor needs to have sufficient temporal resolution so that it can make a significant fit.

Assumption # 10: It is necessary to allow enough overetching so that the tail portion of the endpoint trace appears when the film to be etched is removed.

In summary, the overall suitability assessment and analysis of the end point trace of FIG. 11 is performed sequentially by the following steps (steps 1-15):

A) Measurement before etching: Step 1 Thin film thickness: H = 1913 Å, ± U ₀ = 0.01
(1.0%) Thickness of lower layer: d ₀ = 400 angstrom

B) Observation from end point trace: Step 2 Before end point: S (t) = i ₀ + s ₀ t, where i ₀ = 4861 counts, s ₀ = 0.24 counts / sec Step 3 During end point period: _{S (t) i e + s} e t, where i _e = 54562 counts, s _e = -832 counts / sec step 4 endpoint after: S _∞ = 2000 total etching time including the step 5 overetching of ± 100 count is estimated : T _E = 72.
9 ± 0.05 seconds

C) Calculation: Step 6 t ₁ = (i ₀ −i _e ) / (S _e −S ₀ ) = 59.72 seconds Step 7 t ₂ = (S _∞ −i _e ) / S _e = 63. 18 seconds Step 8 Total etching time to end point: T = (t ₁ + t ₂ ) / 2
= 61.45 seconds Step 9 S ₀ = i ₀ + s ₀ t ₁ , S ₀ = 4875 counts Step 10 δ = 2Ts _e / [(S _∞ −S ₀ )] = 35.561 Step 11 Check δU _n <1. Yes, δU _n = 0.36, pass Step 12 Average etching rate of thin film: E = H / T = 1868 Å / min Step 13 Uniformity of etching rate: u = [− δ + (δ ² +4 (1 + δU _n )] ^1/2 ] / 2 =
0.0381, where u <± 3.8%

D) Post-Etch Measurements: Step 14 Determine the residual thickness of the lower layer with a minimum of 18 measurements across the wafer: d _m = 230 Å.

E) Etching rate calculation: Step 15 Etching rate ratio R _m = E (1-u) (T _E −t ₁ ) / (d ₀ −d _m ) =
2.32, where the selection ratio R can be read as R> 2.32: 1.

When the test wafer having the partially etched pattern was measured using the Wild Lights system under the same etching conditions, the etching rate was 1840 angstrom / min, and the uniformity was ±.
The ratio was 5.78%, and the minimum selection ratio was 1.98: 1. In order to estimate the micro-rotating effect, the measurements before and after the etching were carried out at 3 points for each of the 18 dies (15 × 15 mm), thus a total of 54
Dispersed in different places.

Measurements were performed on the partially etched test wafer under different etching conditions to obtain values for an etch rate of 7260 Å / min, uniformity of ± 11.7% and minimum selectivity of 1.85: 1. Was given. The result of the end point trace in this case is that the etching rate is 6326 Å / min, the uniformity is ± 7.0%,
The selection ratio was 1.7: 1.

Therefore, the data obtained by the end point trace analysis is consistent with the case of the conventional method using the test wafer off-line, but the absolute value is slightly different. No microloading effect was observed, as confirmed by hypothesis # 7.

In the first experiment, the etching rates were very close between the two methods, but in the second experiment they were quite different, in which case the test wafers had a high etching rate of 10 seconds. Only had to etch. In this case, optical techniques are expected to yield useful values.

The difference in the homogeneity values comes from hypothesis # 8.
This is because it did not take into account the start and end of the end point period by estimating the slope (see Figure 11),
It is considered that these areas are included in the method in which 54 points are measured on one wafer using the test wafer.

The tendency of the selection ratio is similar. It should be noted that under these two etching conditions, the minimum selection ratio values are different, but the average values are both 2.1: 1.

The margin of error can be determined for the values of etching rate, uniformity and selectivity based on noise in the radiant intensity signal and known measurement errors.

In general, the feasibility of in-line optical emission intensity analysis during plasma etching depends on the feasibility of the technique for characterizing etch rate and uniformity and selectivity. The method is the same as the above-mentioned 10
It applies when the assumptions of the items are met, but the most important restrictions in terms of importance are i) no significant loading effect, ii) the synchrotron radiation signal at most of the transition period at the end point. It changes linearly, and iii) the thin film to be etched is removed and then sufficiently overetched. Other requirements are
It can be met by careful choice of hardware and wavelengths to monitor. The method of the present invention is applicable to, but not limited to, nitride island etching, polysilicon etching, oxide spacer etching, contact etching.

The method of the present invention is useful in performing an experimental design technique (DOE) in which line width data is collected by comparing several processes. In that case, etching characteristics can be obtained while the wafer is being etched towards its completion. On the other hand, the conventional method requires another special test wafer for partial etching.

By coupling to an in-line analysis module, this method is useful for characterizing the wafer product to be etched, and may be performed independently if desired. Thus, as long as the device operates within the specified specifications, the off-line aptitude evaluation is unnecessary.

[0109]

The technique of the present invention provides a novel method for determining etch rate uniformity during a plasma etch process on a fabrication wafer or test wafer in a single wafer etch system. Under the proper conditions, such as those obtained in today's advanced process equipment, the etch rate and etch uniformity, and in some cases the etch selectivity to the underlying material, is dependent on the optical emission from the plasma itself. It can be derived by analysis of the optical signal. According to this method, in-line suitability evaluation of the etching apparatus can be performed without the need for a test wafer or off-line measurement. Since the etching is performed until completion, data that appears in the final result, such as line width data, can be easily obtained.
In addition, the production of the wafer under test can be resumed without compromising its quality.

In the case of the conventional laser irradiation method,
It requires a laser light source and an irradiation port in an etching device. Another constraint is that the laser is generally only applied to the small spots on the wafer being monitored, thus only local information characterizes the overall wafer result. On the other hand, according to the technique for monitoring the plasma intensity itself according to the present invention, an extra device and an opening are not required, and the entire wafer is monitored. Further, according to the technique of the present invention, it is possible to easily monitor the wafer that is the target of actual production (that is, processing in the actual manufacturing process).

[Brief description of drawings]

FIG. 1 is a schematic view showing optical interference caused by multiple reflection in a transparent thin film on a reflective substrate as a conventional technique.

FIG. 2 is a graph plotting the amount of reflection on an oxide on silicon as a function of oxide thickness, as a relative percentage.

FIG. 3 is a plot (indicated by a square mark) of the relationship between the plasma radiant light intensity (expressed as a “count value”) and the actual etching time (seconds), and the plasma radiant light intensity and the theoretical reflectance. It is the graph which plotted the relationship with (black dot).

FIG. 4 is a graph plotting the relationship between the thickness of a thin film (eg oxide) and the etching time.

FIG. 5 shows a uniform blanket oxide and a non-uniform patterned oxide, 436, respectively, when etched using a rainbow etching apparatus.
FIG. 4 is a graph plotting the relationship between plasma radiation intensity and time obtained from an experiment of monitoring plasma radiation at nm, as in FIG. 3.

FIG. 6 is a flowchart showing each step when the suitability evaluation of the etching process is performed off-line.

FIG. 7 is a flowchart showing each step in the case of performing the in-line evaluation of the suitability of the etching process in real time.

FIG. 8 is two graphs showing a thin film removal pattern for an exposed area “A” in a non-linear manner.

FIG. 9 is two graphs showing a pattern for linear thin film removal for exposed area “A”.

FIG. 10 is four graphs showing simulated end point curves for thin film removal of wafers with a radial symmetric etch rate.

FIG. 11 is a graph showing a typical end point trace in accordance with the present invention.

Claims (9)

[Claims] 1. A method for monitoring the removal of a thin film on a wafer product during plasma etching: monitoring plasma emission light intensity at a selected wavelength; An etching monitoring method in plasma etching, comprising: relating to a residual thin film thickness; 2. The etching monitoring method in plasma etching according to claim 1, further comprising determining a selectivity ratio (thin film: underlying material selectivity ratio) in the etching step. 3. A method according to claim 1, further comprising determining the rate and uniformity of etching based on changes in plasma radiation intensity during etching. 4. The etching monitoring method in plasma etching according to claim 1, wherein the plasma emission light intensity is monitored during the etching of the nitride island portion. 5. The etching monitoring method in plasma etching according to claim 1, wherein the plasma emission light intensity is monitored during etching of polysilicon. 6. A method according to claim 1, wherein the plasma emission light intensity is monitored during etching of the oxide spacer. 7. A method of etching monitoring in plasma etching, wherein the intensity of plasma radiation light is monitored during etching of the contact portion. 8. The method of claim 1, further comprising: etching by a high etching rate and low selectivity etching technique until a predetermined residual thin film thickness is obtained; subsequently at least the thin film is removed. Etching with a low etching rate and a high selection ratio etching method until the end of the etching; 9. The method according to claim 8, further comprising: in order to ensure the removal of the thin film, after the thin film is removed, the etching using the low etching rate and high selectivity etching method is continued. Characterized by
Etching monitoring method in plasma etching.