TWI520245B - Methods of characterizing semiconductor light-emitting devices based on product wafer characteristics - Google Patents

Description

Method for characterizing a semiconductor light emitting device based on product wafer characteristics

The present invention relates to the fabrication of semiconductor light-emitting devices (LEDs) for light-emitting diodes and laser diodes, and more particularly to a method for characterizing semiconductor LEDs based on product wafer characteristics.

In virtually all lighting and lighting applications, semiconductor LEDs such as light-emitting diodes and laser diodes are rapidly replacing conventional light sources. As a result, semiconductor LEDs are being fabricated for a very wide range of emission wavelengths in increasing numbers.

Example semiconductor LEDs for general daylighting are light emitting diodes (also known as "LEDs") and laser diodes. Phosphorescent coatings can be used to create a "white" spectrum of light. The phosphor reacts with a particular emission wavelength of the device (eg, blue wavelength), and Stokes shifts a portion of the emitted light from a shorter wavelength to a longer wavelength to impart an output white spectrum. The white spectrum can be characterized by an equivalent color temperature associated with the emission spectrum of the corresponding blackbody radiation. The "warm" white spectrum is characterized by a color temperature of approximately 2800 °K, while the "cold" white spectrum has a color temperature of approximately 5000 °K. In many applications, warm white spectroscopy is preferred.

To obtain an appropriate color temperature, it is necessary to make the emission wavelength [lambda] of the semiconductor light emitting device _E match absorbed by the phosphor and an emission spectrum of △ λ. Typically, the actual emission wavelength λ _E needs to differ from the desired (selected) emission wavelength λ _ED by no more than +/- 2 nm to properly match the phosphor absorption and emission characteristics. After proper matching, the LED luminaires provide "white light" with a color temperature of approximately 2800 °K. Devices that are outside of a particular wavelength specification have much less value because they produce "discoloration" and are therefore less of a light desired by the consumer. LED manufacturers often sell these "discoloration" LEDs to applications where color is less critical, such as flashlights or external parking facilities. However, the value of such LEDs is much smaller than the LEDs sold to the general household lighting market, where color temperature is critical. To this end, LED manufacturers strive to create more LEDs in each of the more valuable spectral ranges.

For optimum yield and therefore best value and profit, it is desirable to fabricate a semiconductor light emitting device that has an accurate emission wavelength within a specified tolerance. There is also a need to be able to know in advance the emission wavelength of the luminescent wafer (grain) entering the final semiconductor light-emitting device during the manufacturing process.

A method of characterizing a semiconductor light emitting device (LED) based on product wafer characteristics is disclosed. The methods include measuring at least one product wafer characteristic, such as curvature or device layer stress. The method also includes establishing a relationship between the at least one characteristic and an emission wavelength of the LED dies formed from the product wafer. This relationship allows prediction of the emission wavelength of the LED structure formed in the device layer of a similarly formed product wafer. This in turn can be used to characterize product wafers and perform process control in a large number of LED fabrications.

According to one aspect of the invention, a method includes dividing a test product wafer to form a die by measuring a device layer stress S(x, y) of a test product wafer, and measuring the grains The emission wavelength of the device layer is determined by the device layer stress and the corresponding device structure position (x, y) to establish the device layer stress S(x, y) and the emission wavelength of the device structures formed on the product wafer. A relationship between λ _E (x, y). The method further includes measuring a curvature change ΔC(x, y) of the test product wafer based on a curvature measurement of the substrate performed before and after processing the substrate to form the device structures. The method also includes calculating the device layer stress S(x, y) in the test product wafer based on the measured curvature change ΔC. And the method further includes correlating the actual emission wavelength λ _E with the device structure based on the relationship between the device layer stress S(x, y) and the (x, y) position on the test product wafer To establish a predicted emission wavelength of the device structure corresponding to the emission wavelength λ _E .

The method preferably further comprises calculating a stress S(x,y)S(x,y)={M _S h _S ² /6h in the test product wafer based on the measured curvature change ΔC via the following relationship _F }△C(x,y)

Where M _{s is a} biaxial modulus of the substrate 20, h _{S is a} height of the substrate, and h _F is a thickness of one of the device structures.

In this method, the device structures can have a light emitting diode or a laser diode configuration.

The method preferably further includes establishing a desired emission wavelength, dividing the product wafer to form a die from the device structures, and selecting the grains based on the predicted emission wavelength.

The method preferably further includes defining a tolerance of the amount of change in the emission wavelength, and comparing the predicted emission wavelength to the tolerance, and determining an action scheme from the comparison, including discarding the product wafer, heavy Process the wafer of the product or use a selected portion of the wafer of the product.

In the method, the substrate processing can include performing an organometallic chemical vapor deposition (MOCVD) process.

Another aspect of the invention is a method of characterizing one of the semiconductor light emitting device (LED) structures of a product wafer having a substrate and a device layer. The method includes measuring at least one characteristic of a test product wafer in substantially the same manner as the product wafer, wherein the at least one characteristic is selected from the group consisting of: a device layer stress and a product wafer curvature characteristic . The method also includes dividing the at least one test product wafer to form LED dies having a location associated with the test product wafer associated therewith. The method also includes measuring an LED emission wavelength of each of the LED dies to establish a set of LED emission wavelengths as a function of position on the product wafer. The method further includes determining a relationship between the wafer characteristics of the at least one test product, the set of LED emission wavelengths, and the locations of the LED dies, and predicting the relationship using the relationship determined in the action "decision-relationship" The LED emission wavelength of the LED structures on the product wafer.

The method preferably includes sorting the LED structures of the product wafer based on the relationship in the action "Decision-Relationship".

In the method, the product wafer preferably includes a substrate. And the method preferably further comprises measuring a curvature of the substrate prior to forming the product wafer.

The method preferably further includes measuring at least one of a substrate curvature and a product wafer curvature using coherent gradient sensing (CGS).

In this method, the device structures can have a size. And the method is preferably The determining at least one test product wafer characteristic is determined at a size substantially equal to or less than one of the device structural dimensions.

Another aspect of the invention is a method of forming a semiconductor light emitting device (LED). The method includes forming a product wafer comprising an LED structure formed on a substrate by one of a process variable having one or more process variables, wherein the substrate has a known initiality prior to forming the LED structures on the substrate Curvature C ₀ (x, y). The method further includes measuring a curvature C(x, y) of the product wafer after forming the light-emitting device structure on the product wafer, and determining a curvature change ΔC(x, y)=C(x, y) - C ₀ (x, y). The method further includes calculating a stress S(x, y) in the product wafer based on the measured curvature change ΔC(x, y). The method further includes causing one of the LED structures to emit a wavelength based on a relationship between the calculated stress S(x, y) and the (x, y) position of the illuminator structures on the product wafer λ _{E is} associated with the calculated stress S(x, y). And the method further includes comparing the emission wavelength λ _E to an emission wavelength variation tolerance, and sorting the LED structures into one or more bins based on the tolerance.

The method preferably further includes forming the LED device using only the LED structures in one or more selected bins.

The method preferably further includes adjusting at least one of the one or more process variables to reduce the structure of the illuminating devices on the product wafer in the illuminating device structure (x, y The amount of change in the wavelength of the change in position.

In the method, the process variables are preferably selected from processes comprising the following A collection of variables: time, temperature, temperature uniformity, gas partial pressure, gas partial pressure uniformity, gas flow rate, and gas flow uniformity.

The method preferably further includes performing a curvature measurement of one or more test product wafers in the same manner as the product wafer, dividing the one or more test product wafers to form LED dies, and Detecting the emission wavelengths of the LED dies for the (x, y) position of the LED dies on the one or more test product wafers a function.

The method preferably further includes forming a new product wafer, measuring the curvature of the new product wafer, and determining an LED structure on the new product wafer based on the measured curvature of the new product wafer LED emission wavelength.

Another aspect of the invention is a method of forming a semiconductor light emitting device (LED). The method includes forming a product wafer including an LED structure formed on a semiconductor substrate by one of a process variable having one or more process variables, and measuring the product wafer after forming the LED structures on the product wafer One of curvature uniformity, and adjusting at least one of the one or more process variables to meet at least one of a curvature uniformity requirement and a stress uniformity requirement.

Another aspect of the invention is a method of forming a semiconductor light emitting device (LED). The method includes forming a product wafer comprising an LED structure formed on a substrate by one of a process variable having one or more process variables, wherein the substrate has a known initiality prior to forming the light emitting device structure on the substrate curvature C ₀ (x, y). The method also includes measuring a curvature C(x, y) of the product wafer after forming the light-emitting device structure on the product wafer, and determining based on C ₀ (x, y) and C (x, y) A uniformity of curvature. The method also includes determining a first number of grains belonging to a first range of curvature uniformity. The method also includes determining a second number of grains belonging to a second range of curvature uniformity. And the method also includes assigning a quality value to the product wafer based on the first number and the second number.

The method preferably further includes arranging the product wafer based on the assigned quality value.

The method preferably further includes splitting the product wafer to form LED dies associated with the first number and the second number. The method preferably further includes using the LED dies associated with the first number for a first application, and using the LED dies associated with the second number for a second application.

Another aspect of the invention is a method of forming a semiconductor light emitting device. The method includes forming a product wafer comprising a light-emitting device structure formed on a substrate by one of a process variable having one or more process variables, wherein the substrate has a known structure before the light-emitting device structure is formed on the substrate Initial curvature C ₀ (x, y). The method also includes measuring a curvature C(x, y) of the product wafer after forming the light-emitting device structure on the product wafer, and determining a curvature change ΔC(x, y)=C(x, y) - C ₀ (x, y). The method also includes causing one of the illuminating device structures to emit a wavelength based on a relationship between the calculated curvature C(x, y) and the (x, y) position of the illuminating device structures on the product wafer λ _{E is} associated with the calculated wafer curvature C(x, y). And the method also includes comparing the emission wavelength λ _E with an emission wavelength variation tolerance to determine which illumination structures can be used to form a illumination device.

The additional features and advantages of the invention are set forth in the description which follows, and in part Figure) to recognize. The scope of the invention is incorporated in the Detailed Description of the Invention and is a part of the detailed description of the invention.

It is to be understood that both the foregoing general description and the following description of the embodiments of the invention The accompanying drawings are included to provide a further understanding of the invention The drawings illustrate various embodiments of the invention and, together,

Reference will now be made in detail to the preferred embodiments embodiments Wherever possible, the same reference numerals and the The scope of the patent application is incorporated into and constitutes a part of this detailed description.

In the following discussion, the device layer stress is the stress in the device layer 30 that defines the device 50 formed in the device layer 30, and is represented by S(x, y) or only S, which is S(x, y; σ _{i , j} ) shorthand notation, where σ _i,j is the stress tensor. The stress tensor σ _i,j defines the stress at any point in the device layer of the substrate. The coordinates (x, y) represent the wafer Cartesian coordinates of the point in the device layer 30 where the stress is defined. The stress tensor σ _i,j defines three orthogonal normal stress components (σ ₁₁ , σ ₂₂ , σ ₃₃ ) and six orthogonal shear stresses (σ ₁₂ , σ ₁₃ , σ ₂₁ , σ ₂₃ , σ ₃₁ , σ ₃₂ ). When the coordinate system used to measure stress is oriented in the main direction, the stress tensor contains only three orthogonal normal stress components. In one example, S(x,y) has a single value representative of stress, and this single value can be, for example, a maximum stress component, a sum of stress components, an average of stress components, or some combination of stress components. In another example of using a rate of stress change at a given (x, y) position, the notation S'(x, y) is used.

Also, in the following discussion, the curvature C(x, y) of the surface is a shorthand for C(x, y; n ), where n is the normal vector in a given direction in the XY plane, and it is defined with the XY plane A plane intersecting the surface at right angles, and C(x, y) is the curvature of the curve formed by the surface when intersecting the plane. Curvature is generally defined as 1/R, where R is the local radius of the curvature of the surface at a point (x, y) in the plane defined by the normal vector n . In one example, n is taken in a single direction, in which case C(x, y) is a scalar function. Typically, C(x,y) is a tensor because it can have multiple curvatures associated with it, i.e., one curvature for each direction of the unit vector n . In one example, the height of each (x,y) surface point can be customized (as measured relative to a reference plane (eg, a perfectly flat wafer) by a function H(x, y) to determine the curvature C(x, y). In detail, H(x) can be obtained by fitting a circular curve to a point around a point (x, y) in a given plane to obtain a best fit radius R and then determining the curvature as 1/R. , y) obtain C(x, y). C(x, y) can also be obtained from the second derivative of H(x, y) using standard mathematical techniques. The notation C _{0 is} used to indicate the curvature of the substrate used to form the product wafer.

In the discussion herein, the abbreviation "LED" is generally understood to mean "lighting device", but it can also mean "light emitting diode", and those skilled in the art will understand based on the context of using this abbreviation. The difference.

1 is a plan view of an example product wafer 10 used to form a semiconductor LED, such as a light emitting diode and a laser diode, and FIG. 2 is a cross-sectional view thereof. As used herein, the term "product wafer" generally means a wafer or substrate having a device structure formed thereon, and wherein the device structure is then used to fabricate a light emitting device, in one example, the light emitting device can be used to form an LED product or device. .

The product wafer 10 includes a semiconductor substrate 20 having an edge 21, a top surface 22, and a bottom surface 24, wherein the device layer 30 is formed on the surface 22. An example semiconductor substrate 20 is made of sapphire or germanium. Device layer 30 includes an array 32 of semiconductor light emitting device structures ("device structures") 40. In one example, product wafer 10 includes thousands of device structures 40 that can have a size of about 1 mm. For the sapphire substrate 20, the example product wafer 10 has a diameter of 2 吋 to 6 , and has a diameter of 6 吋 to 12 矽 for the 矽 substrate. The device structure 40 has an actual emission wavelength λ _E associated therewith and an output spectrum Δλ associated with the aforementioned color temperature. The desired or selected emission wavelength of device structure 40 is expressed as λ _ED .

Once the product wafer 10 is fully processed such that the device structure 40 functions, the product wafer 10 is diced ("split") such that individual device structures in the array 32 are separated to form individual dies 42. The die 42 is then incorporated into a semiconductor light emitting device structure to form a light emitting device 50, such as illustrated in Figure 3 as a perspective view of an example LED device. The LED device 50 of FIG. 3 includes an anode 52A and a cathode 52C that extend into the interior 54 of the epoxy lens housing 56. Cathode 52C includes a reflective cavity 58 in which die 42 is located. Wire bond 60 electrically connects anode 52A and cathode 52C to the LED. A power source (not shown) is coupled to anode 52A and cathode 52C to provide the power required to power the LED such that the LED emits light 62 at the actual wavelength λ _E . When the die 42 is a prior art die, the LED device 50 of FIG. 3 is a prior art device, but when the die 42 is formed using the methods disclosed herein, the LED device 50 of FIG. 3 is not a prior art device.

The example device structure 40 is in the form of an LED fabricated by growing GaN on a sapphire substrate 20. GaN is grown using a metalorganic chemical vapor deposition (MOCVD) process. The MOCVD process is performed in an MOCVD reactor and is performed in a manner that results in the formation of a multi-quantum well structure (not shown).

4 shows a collection of sapphire substrates 20 supported in a substrate holder or wafer holder 70 having a top surface 72. FIG. 5 is a close-up cross-sectional view of the crystal holder 70 illustrating an example of the manner in which the substrate 20 is supported in the crystal holder 70. The crystal holder 70 includes an opening 76 formed in the top surface 72, wherein the opening defines an end edge 78 that supports the substrate 20 near the edge 21 of the substrate on the bottom surface 24 of the substrate.

FIG. 6 shows an MOCVD reactor system 90 having a reactor chamber 100 operatively coupled to an MOCVD subsystem 96. The MOCVD subsystem 96 includes various MOCVD system components (not shown) such as a vacuum pump, a gas source, a ventilation system, a baratron (film capacitive pressure sensor), and the like. The MOCVD chamber 100 has The crystal holder 70 and the substrate 20 are located inside the interior 104 when the MOCVD process is performed. Note that the substrate 20 becomes the product wafer 10 after the MOCVD process is performed. MOCVD reactor system 90 includes a controller 110 that is operatively coupled to MOCVD subsystem 96. Controller 110 is configured to control the MOCVD process occurring in reactor interior 104 (as indicated by arrow 105).

In particular, the controller 110 is used to carefully control the temperature T of the product wafer 10 because the actual emission wavelength λ _E varies significantly with MOCVD growth conditions. 7A-7D are plan views of an example product wafer 10, which schematically illustrate temperature T(x, y), device size D(x, y), device layer stress S(x, y), and (actual), respectively. An example contour of the emission wavelength λ _E (x, y). Here, since the substrate 20 is undergoing processing to form the product wafer 10, the temperature T refers to the substrate temperature or the product wafer temperature. For ease of discussion, the following discussion refers to "product wafer temperature."

As the product wafer temperature T(x,y) changes (Fig. 7A), the device size D(x,y) (eg, the thickness of a multiple quantum well structure (not shown) formed during the MOCVD process) is in a corresponding manner Change (Figure 7B). This is also translated as a corresponding change in the device layer stress S(x, y) on the product wafer 10 (Fig. 7C), which in turn results in a corresponding change in the actual emission wavelength λ _E (x, y) (Fig. 7D).

In some cases, the product wafer temperature T 1 ° K temperature change may result in the emission wavelength λ _E of approximately 1 nm displacement δλ _E. Therefore, it is desirable to first detect and ultimately control temperature non-uniformity and temperature repeatability in the MOCVD reactor system 90 to ensure proper control of the product wafer temperature T. Temperature non-uniformity on the product wafer can result in localized changes in growth conditions that can result in changes in the LED emission wavelength.

In today's MOCVD-based fabrication of device structure 40, the actual emission wavelength λE and corresponding emission uniformity are up to the device wafer 40 from the product wafer 10 to form the die 42 and then the die is incorporated into an LED such as FIG. The illumination device or equivalent test structure of device 50 is known.

Currently, in order to estimate or predict the LED emission wavelength λ _E , the substrate is inspected by photoluminescence techniques in which a short wavelength source (typically 248 nm) is incident on the multiple quantum well region to excite the emission. However, this technology is significantly limited by its point-by-point inspection technique. Accurate mapping of the entire product wafer by high spatial resolution (eg, less than the spatial resolution of the grain size) takes from 30 minutes to 240 minutes, depending on the substrate size used to form the product wafer.

A further limitation of this technique is that the emission wavelength from photoluminescence is typically different from the emission wavelength from the LED during electrical stimulation. The root cause of this difference is believed to be the additional manufacturing steps experienced by the LED between the photoluminescence inspection and the final product. Typically, there is a pre-measured deviation in production between the photoluminescence emission wavelength and the LED emission wavelength of the electrical stimulus. This is then used as a process monitor.

The amount of device layer stress S(x, y) in the device structure 40 formed on the product wafer 10 and the history of the product wafer temperature T(x, y) (ie, T(x, y, t ), Where t is time) directly related. Fundamentally, the difference in product wafer temperature during the deposition process (i.e., non-uniform heating) results in a thickness variation in the deposited GaN layer (and in particular, the multi-quantum well layer). After deposition, and when the substrate is returned to room temperature, the difference in thermal expansion coefficients of the constituent materials of the constituent device structures 40 results in relative expansion or contraction of the device structure 40 relative to the substrate 20 during thermal cycling. Non-uniform heating results in non-uniform device layer stress and non-uniform emission wavelength (λ _E ).

This relative expansion or contraction causes stress in the device structure 40 formed on the product wafer 10 such that thermal energy imparted to the product wafer is partially stored as mechanical energy in the device layer 30. The non-uniformity (i.e., variation) of the product wafer temperature T(x, y) during deposition results in layer thickness and non-uniformity (change) in the final device layer stress S(x, y). A product wafer 10 having uniform but different temperatures T(x, y) will have a uniform but different device layer stress S(x, y).

The device layer stress S(x, y) can be monitored via the change in the curvature of the substrate ΔC(x, y). The most common stress metering systems rely on measurements of changes in the shape of the product wafer to calculate device layer stress. The device stress S(x, y) in the device structure 40 supported by the product wafer 10 is calculated using the Stoney formula from the curvature change ΔC(x, y): S(x, y) = {M _S h _S ² / 6h _F }△C(x,y) Equation 1

Where M is the biaxial modulus, h is the thickness, and the subscript "F" represents the deposited film (GaN in this example), and the subscript "s" represents the substrate (sapphire or germanium in this example). . The change in C(x, y) on the wafer indicates a non-uniform device layer stress S.

After knowing that the temperature and thermal expansion coefficient (α) mismatch between the film (ie, the device layer 30) and the underlying substrate 20 does not match to cause the device layer stress S, we can derive the stress S and curvature for the device layer. The following expression of C: S(x, y) = M _f ε _m (T) = M _f (α _{f -} α _s ) ΔT Equation 2

C(x,y)={6M _f h _f /M _s h _s ² }(α _f -α _s )ΔT Equation 3

Wherein ε _m is the mismatch or misfit strain between the membrane and the substrate, △ T is the stress-free temperature T _o (i.e., the substrate temperature during the film match) the temperature difference.

The above equation shows how the curvature change ΔC(x, y) is related to the change in deposition temperature. Therefore, from these expressions, we can make the product wafer curvature temperature-dependent, and the temperature is directly related to the emission wavelength λ _{E of the} LED.

The equations provided above are based on specific assumptions regarding the stress and deformation state, the geometry of the device structure, and the thermomechanical properties of the constituent materials. A more complex form or version of this equation can be produced for a film/substrate system with any attributes. However, for any particular device/substrate system, an appropriate relationship can be created along the lines set forth herein to correlate substrate curvature changes with temperature changes and emission wavelength variations.

One aspect of the present invention enables at least one product wafer characteristic (such as product wafer stress (i.e., one or more components of stress tensor) and curvature) and at least one LED performance characteristic (such as transmit power, efficiency) , wavelength, spectral bandwidth, and the like) are related. In one example, this is accomplished by empirically establishing a relationship between at least one measured performance characteristic of the LED and at least one product wafer characteristic. In one example, this can be accomplished by evaluating at least one performance characteristic of the LEDs obtained from the product wafer against individual process steps used to form the product wafer. This can also be accomplished by evaluating at least one LED performance characteristic against a plurality of process steps used to form the product wafer.

Accordingly, one aspect of the present invention includes a method of performing quality control using at least one of product wafer stress and product wafer curvature as a process monitor for sorting or selecting individual wafers of a product wafer Performance. The method can also include monitoring deposition characteristics during product wafer formation and performing process optimization. The method can further include matching the performance of a set of two or more of the same type of process tool such that the product wafers match when formed on different process tools and thus their LED dies also match. That is, the product wafer formed by the different process tools in the matching set is similar to the product wafer formed without the process tool.

8A-8C are an example product wafer 10 illustrating an example contour of curvature C(x, y), device layer stress S(x, y), and actual emission wavelength λ _E (x, y). Floor plan. One aspect of the invention relates to determining the device layer stress S(x, y) by measuring the product wafer curvature C(x, y), and then using the determined device layer stress S(x, y) to predict the product The actual emission wavelength λ _E of the resulting (separated) die 42 of the wafer 10 corresponding to the location (x, y) of the device structure 40.

In Figure 8C, a particular device structure 40 is shown having a position (x _i , y _j ) measured relative to a reference location (e.g., the center of the product wafer). Close-up device 32 configuration of the array 40 by 1 nm increments illustration shows actual emission wavelength λ _E of more detailed (i.e., closely spaced) contour λ _E. In one example, an emission wavelength change (offset) δλ _{E of} 1 nm can be predicted.

Figure 8D is a plan view of a hypothetical product wafer 10 showing an example 2 nm contour of the predicted emission wavelength λ _EP with the product wafer 10 having a concave or bowl-shaped curvature C(x, y). Assuming a desired emission wavelength of λ _ED = 456 nm with an emission wavelength variation tolerance δλ of +/- 1 nm, the predicted wavelength contour of Figure 8D allows us to split the product wafer 10 and use only the final illumination device 50 from There are crystal grains 42 of the device structure 40 at the (x, y) position within the 455 nm and 457 nm predicted wavelength contours shown in solid lines (i.e., within the emission wavelength variation tolerance δλ). Thus, in one example, the first number of LED structures 40 are in the range of 455 nm and 457 nm contours, and the second number of LED structures 40 are outside of the contour range. This type of selection of LED structures 40 based on selected specifications (here, emission wavelengths) can be used to arrange the resulting LED dies 42 for the first and second applications or the first and second types of LED devices.

At this point, the number of grains 42 within the acceptance wavelength change (i.e., δλ = +/- 1 nm in this example) can be estimated. In addition, the number of dies 42 that are slightly outside the window but still suitable for "discoloration" applications can be estimated. In addition, the number of dies 42 that are outside of any applicable range and therefore have no intrinsic value can also be estimated. The intrinsic value of the product wafer 10 can be determined based on the number of dies 42 in each wavelength variation range and after knowing the value of the dies 42 for each wavelength variation range. This allows the LED manufacturer to determine if there is a good reason to continue processing the particular product wafer 10. For example, if the die array is outside of a wavelength range of any value, there is little reason to continue processing the wafer 10 and incur additional costs.

In summary, a change in the deposition temperature during LED formation results in a wavelength shift of the emission wavelength λ _E of the LED. These same temperature changes result in a change in stress in the deposited film, which results in a change in wafer curvature. Thus, via the above equation, the change in wafer curvature C can be related to the wavelength shift.

In practice, it may be difficult to know the exact values of the material constants (such as the coefficient of thermal expansion α and the biaxial modulus M) of various materials at high temperatures used during the MOCVD process. For example, typical temperatures in MOCVD processes are often about 1000 °C. Analytically correlating the above equations to the actual LED wavelength shift requires accurate knowledge of these material constants and the temperature history of the substrate 20. This information may not be available.

The necessary information is thus obtained empirically, and by accurately measuring the curvature C of all points on one or more test product wafers 10 and measuring the LED emission wavelength λ _E of the LED dies 42 obtained from the product wafer 10. It may be more convenient to generate a lookup table or correlation curve for the actual wavelength offset. This method allows us to correlate the measured substrate curvature C with the actual device data and to use the substrate curvature as a process control and inspection monitor for subsequent mass LED fabrication.

The main advantage of using the wafer curvature C to monitor the emission wavelength λ _E and emission uniformity over the currently used photoluminescence technology is that the wafer curvature can be measured at very high spatial frequencies in a very short time. A coherent gradient sensing system (described below) can easily sample a substrate by spatial sampling of a few hundred microns (which is typically less than 1 mm) and can be inspected (measured) by approximately 200 mm in approximately 1 minute. Wafer. In comparison, it would take several hours to achieve similar spatial information on a 200 mm wafer using a photoluminescent system.

9 is a schematic diagram of an example coherent gradient sensing (CGS) system 200 that can be used to measure the curvature C(x,y) of a product wafer 10. Details of the manner of operation of the CGS sensing are described in U.S. Patent No. 6,031,611 (the ' 611 patent) which is incorporated herein by reference. Figure 9 is based on Figure 1 of the '611 patent.

The CGS system 200 includes a digital camera 210, a filter lens 224 along the axis A1 (e.g., a filter coupled to the lens, as discussed in the '611 patent and shown in Figure 1 therein), the shaft The first grating G1 and the second grating G2, a beam splitter 230, and a wafer stage 240 are spaced apart. The CGS system 200 also includes a laser 250 disposed along an optical axis A2 that intersects the axis A1 at the beam splitter 230. The CGS system 200 also includes a controller or signal processor 260 that is operatively coupled to the digital camera 210 and to one of the lasers 250. An example signal processor 260 is or includes a computer having a processor 262 and a computer readable medium ("memory") 264, the processor and computer readable medium being configured to control CGS via instructions recorded thereon The operation of system 200 is performed to perform a measurement of the product wafer curvature C(x, y) according to the method described in the '611 patent.

In operation, laser 250 produces a collimated laser beam 252 that is directed by beam splitter 230 to product wafer 10. The collimated laser beam is reflected from the product wafer 10 (and, in particular, from the device layer 30) as reflected light 252R, which travels upward through the beam splitter 230 and through the grating G1 and G2. The two gratings G1 and G2 are spaced apart and otherwise configured to shear the collimated laser beam. Then make The light passing through the gratings G1 and G2 is focused onto the digital camera 210 by a filter lens 224.

The shearing and filtering process produces stripes at the digital camera 210 that are contours of a constant surface slope on the product wafer 10. Thus, two sets of orthogonal interferograms are used to characterize the surface slope, which can be numerically integrated to reconstruct the substrate configuration or can be numerically differentiated to obtain the product wafer curvature C(x,y). Two camera configurations for the CGS system, such as shown in Figure 6 of the '611 patent, can also be used to facilitate the acquisition of the two sets of orthogonal interferograms.

A curvature map (i.e., product wafer curvature C(x, y)) is then generated from the interference fringe pattern using, for example, the method described in the '611 patent as performed in controller 260. In one example, the measured curvature C(x,y) of the product wafer 10 has a spatial resolution of between about 100 microns and 300 microns. Thus, for a device structure 40 having a dimension d (see FIG. 3) of about 1 mm, there may be at least one curvature data point per device, and in some cases there may be multiple data points per device (grain). In one example, the method disclosed herein includes determining a spatial variation in stress S(x, y) in a product wafer at a size substantially equal to or less than the dimension d of the device structure 40.

In one example, the curvature C ₀ (x, y) of the substrate 20 is measured prior to processing the substrate to form the product wafer 10. This allows the calculation of the curvature change ΔC(x, y) = C(x, y) - C ₀ (x, y).

Once the substrate curvature C ₀ (x, y) and the product wafer curvature C(x, y) are measured and the curvature change ΔC(x, y) is calculated, the device layer stress S(x, y) can be determined. In one example, this is accomplished using the Stoney equation (Equation 1 above) relating the device layer stress S to the curvature change ΔC.

For most product wafers 10, the curvature change ΔC(x, y) varies across the wafer and also varies with orientation, suggesting the presence of non-uniform device layer stress S(x, y). Or, if the material properties at different temperatures are not known, we can measure the wavelength offset δλ on the sampling substrate (wafer) versus the measured curvature C, and link these measurements together. Future measurements on product wafer 10.

Other methods can be used to perform the measurement of the allowable decision C(x, y). Such methods include, for example, contour mapping, interference measurements, capacitance gauges, and laser beam deflection. Similarly, the device layer stress S(x, y) can be determined by measuring the distortion of the crystal lattice using, for example, X-ray diffraction and Raman spectroscopy.

Thus, one aspect of the method of the present invention includes the following.

1. Before Before treatment, e.g., during the MOCVD, the shape (curvature) of the measurement substrate 20 C ₀ (x, y).

2. After processing, measure the shape C(x, y) of the product wafer 10.

3. Calculate ΔC(x, y) = C(x, y) - C ₀ (x, y) to determine the curvature change associated with the product wafer 10 induced by the processing substrate 20 to produce the product wafer ΔC .

4. Calculate the device layer stress S(x, y) in the product wafer using the Stoney formula or related formula (eg, Blake's Equation) to calculate the product wafer based on ΔC(x, y) Process induced device layer stress S(x, y).

5. Associate the device structure 40 at the location (x _{i ,} y _j ) with the device layer stress S(x _{i ,} y _j ).

6. Associate the device layer stress value S(x _{i ,} y _j ) with the emission wavelength λ _E (x _{i ,} y _j ), which in turn allows measurement of the product wafer curvature to predict the formation of the LED when the product wafer is segmented The emission wavelength of the LED structure 40 of the die.

The above step 6 requires establishing a relationship between the actual emission wavelength λ _E , the curvature change ΔC and the change in the device layer stress S(x, y) for a given type of process and the device structure 40 being produced. This can be done if all material properties are well known and if the relationship between temperature and deposition rate is well known. As discussed above, such quantities may not be well known and it is necessary to empirically generate data to correlate the change in curvature of the measurement to the change in emission wavelength.

In one example, the empirical "fingerprinting" process can be performed as follows: 1. For a particular process for forming device structure 40, steps 1 through 4 are performed for one or more test product wafers; 2. Measurement The actual emission wavelength λ _E of the LED die obtained from the test product wafer 10 as a function of the device structure position (x, y) is established to establish λ _E (x, y); and 3. the device layer stress S (x, y) The relationship between the actual emission wavelength λ _E (x, y).

The method of the present invention also includes performing process monitoring and process control of the product wafer 10 to determine (predict) the actual emission wavelength λ _E that the LED die 42 will produce when the product wafer 10 is segmented. This step can be performed between different product wafers for a given product wafer and for the same process (ie, the same device structure). Selected LED dies 42 characterized using the methods disclosed herein can then be used in a selected type of LED device 50 such as that shown in FIG.

The method of the present invention also includes performing process optimization for forming device structure 40 by adjusting at least one of process variables used in forming product wafer 10. This includes, for example: 1. Establishing process variables (eg, temperature, temperature uniformity, gas partial pressure, gas partial pressure uniformity, flow rate, flow rate uniformity, time) and process induced stress characteristics (eg, average The relationship between stress and stress uniformity; and 2. modifying at least one of the process variables to optimize the stress characteristics of the process leading to the desired device characteristics, such as the emission wavelength as much as possible The ground is close to the desired emission wavelength.

The method of the present invention, which is intended to optimize the process, also includes the following steps:

1. Establish stress characteristics associated with different processing tools (eg, different MOCVD reactor systems) used to perform the same process.

2. Identify specific processing tools that result in the most undesirable stress characteristics (i.e., the stress characteristics that result in the greatest amount of device performance variation (e.g., the largest change in actual emission wavelength).

3. Identify processing tool parameters such as hardware or control settings, adjustments, etc. that affect stress characteristics (ie, wafer temperature, grain temperature uniformity, gas partial pressure, gas partial pressure uniformity, gas flow rate, gas flow) Rate uniformity, etc.).

4. Adjusting the processing tool parameters to obtain a change in the reduced emission wavelength λ _E , and in a particular example, minimizing variations in the emission wavelength λ _E of the LED dies.

10 is a flow chart 300 illustrating an example method for predicting an actual emission wavelength λ _E of an LED die 42 formed from a product wafer 10 based on a curvature measurement of a product wafer and a substrate used to form the product wafer.

Flowchart 300 includes a first step 302 of establishing a relationship between device layer stress S(x, y) and the actual emission wavelength λ _E (x, y) of die 42 formed from device structure 40 on product wafer 10. .

Flowchart 300 includes a second step 304 that measures the curvature change ΔC(x, y) of product wafer 10 based on substrate curvature measurement C ₀ (x, y) (as described above).

Flowchart 300 includes a third step 306 of calculating the device layer stress S(x, y) of the product wafer 10 based on the measured curvature change ΔC using, for example, the Stoney equation.

Flowchart 300 includes a fourth step 308 of correlating the (x,y) device structure 40 on the product wafer 10 to form the die 42 with the calculated device layer stress S(x,y).

Flowchart 300 includes a fifth step 310 of associating actual emission wavelengths λ _E with various dies 42 based on the device layer stress values S(x, y) established in step 302.

Flowchart 300 includes a sixth step 312 of adjusting at least one process variable to reduce the change in emission wavelength λ _E of die 42.

In another example of the methods disclosed herein, product wafer curvature measurements can be performed at different stages (steps) of completion of product wafer processing. This provides insight into how the product wafer curvature changes for each process step.

Again, as discussed above, flowchart 300 can include a sub-step as part of step 302 that empirically obtains the necessary data from test product wafer 10 to accurately measure the curvature on the sample (set) of the wafer C and later measuring its LED emission wavelength λ _E generated wavelength shift of the actual correlation curve or lookup table so that the amount of the substrate measured curvature C means associated with the actual data. This allows the product wafer curvature measurement to be used as a process control and inspection monitor for subsequent mass LED device fabrication.

It will be apparent to those skilled in the art that various modifications and changes can be made in the present invention without departing from the spirit and scope of the invention. Therefore, it is intended that the present invention cover the modifications and variations of the invention, as the scope of the appended claims and their equivalents.

10‧‧‧Product Wafer

20‧‧‧Substrate

21‧‧‧ edge

22‧‧‧ top surface

24‧‧‧ bottom surface

30‧‧‧Device level

32‧‧‧Array

40‧‧‧Semiconductor light-emitting device structure ("device structure")

42‧‧‧ grain

50‧‧‧Lighting device/LED device

52A‧‧‧Anode

52C‧‧‧ cathode

54‧‧‧Internal

56‧‧‧Epoxy lens housing

58‧‧‧Reflection chamber

60‧‧‧ wire bonding

62‧‧‧Light

70‧‧‧crystal seat

72‧‧‧ top surface

76‧‧‧ openings

78‧‧‧Edge

90‧‧‧MOCVD reactor system

96‧‧‧MOCVD subsystem

100‧‧‧MOCVD chamber

104‧‧‧Reactor inside

105‧‧‧ arrow

110‧‧‧ Controller

200‧‧•Coherent Gradient Sensing (CGS) System

210‧‧‧Digital camera

224‧‧‧ Filter lens

230‧‧‧beam splitter

240‧‧‧wafer stage

250‧‧ ‧ laser

252‧‧‧ collimated laser beam

252R‧‧‧reflected light

260‧‧‧Controller or signal processor

262‧‧‧ processor

264‧‧‧ Computer-readable media ("memory")

300‧‧‧ Flowchart

A1‧‧‧ axis

A2‧‧‧ optical axis

G1‧‧‧ first grating

G2‧‧‧second grating

1 is a plan view of an example product wafer used in the formation of a semiconductor light emitting device, and FIG. 2 is a cross-sectional view thereof; FIG. 3 is a view showing a manner in which a die from a product wafer is incorporated into a structure of an LED device. Figure 4 shows a collection of sapphire substrates supported in a crystal holder; Figure 5 is a close-up cross-sectional view of a portion of the crystal holder showing the manner in which the substrate is supported in the crystal holder; Figure 6 A schematic diagram of an MOCVD reactor system for forming a device layer on top of a substrate to form the product wafer of FIG. 2; FIGS. 7A-7D are plan views illustrating product lines of the contours, respectively, showing the product Variations of wafer temperature T, device size D, device layer stress S, and emission wavelength λ _E ; FIGS. 8A-8C are plan views illustrating product lines of contours showing product wafer curvature C, respectively , device layer stress S and emission wavelength λ _E change; FIG. 8D is a product wafer similar to FIG. 8C and illustrates a plan view of a product wafer showing an example of a predicted emission wavelength, and wherein 457 nm and 455 nm are equal. The value line is bold, this is because It represents the boundary of the device structure with the desired wavelength of 456 nm +/- 1 nm; Figure 9 is a schematic diagram of an example coherent gradient sensing (CGS) system for measuring the curvature of the substrate and product wafer in one example. And Figure 10 is a flow chart illustrating an example method for predicting the emission wavelength of a die from a product wafer based on the curvature measurement of the product wafer.

For reference reasons, Cartesian coordinates are shown in some figures, and such coordinates are not intended to be limiting as to orientation or configuration.

10‧‧‧Product Wafer

Claims (22)

A method of characterizing an emission wavelength of a semiconductor light emitting device structure of a product wafer having a substrate, comprising: dividing the test by measuring a device layer stress S(x, y) of a test product wafer The product wafer is formed into crystal grains, and the thickness of the crystal grains is determined by the device layer stress and the corresponding device structure position (x, y) to establish the device layer stress S (x, y) ) formed on a relationship between the emission wavelength of the device structures of these the product wafer λ _E (x, y); based on the substrate prior to forming the substrate processing apparatus and the configuration of those executed after a curvature measurement to measure a curvature change ΔC(x, y) of the test product wafer; and based on the measured curvature change ΔC, calculate the device layer stress S(x, in the test product wafer y); and based on the relationship between the device layer stress S(x, y) and the (x, y) position on the test product wafer, the actual emission wavelength λ _{E is} associated with the device structure To establish a predicted emission wavelength of the device structure corresponding to the emission wavelength λ _E . The method of claim 1, further calculating the stress S(x, y) S(x, y)={M _S h in the test product wafer based on the measured curvature change ΔC by the following relationship _S ² /6h _F }ΔC(x,y) wherein M _{s is a} biaxial modulus of the substrate, h _{S is a} height of the substrate, and h _F is a thickness of one of the device structures. The method of claim 1, wherein the device structures have a light emitting diode or a laser diode configuration. The method of claim 1, further comprising: establishing a desired emission wavelength; dividing the product wafer to form a die from the device structures; and selecting the grains based on the predicted emission wavelength. The method of claim 1, further comprising: defining a tolerance of a change in the one of the emission wavelengths; and comparing the predicted emission wavelength to the tolerance, and determining an action plan from the comparison, comprising: Discard the product wafer, rework the product wafer, or use a selected portion of the product wafer. The method of claim 1, wherein the substrate processing comprises performing an organometallic chemical vapor deposition (MOCVD) process. A method of characterizing a wavelength of a semiconductor light emitting device (LED) structure of a product wafer having a substrate and a device layer, comprising: a) measuring a test in the same manner as the product wafer At least one characteristic of the product wafer, wherein the at least one characteristic is selected from the group consisting of: a device layer stress and a product wafer curvature; b) dividing the at least one test product wafer to form an LED die, the LEDs The die has a location associated with the test product wafer; c) measuring an LED emission wavelength of each of the LED dies to determine a set of LED emission wavelengths that vary with position on the wafer of the product; d) determining a relationship between the wafer characteristics of the at least one test product, the set of LED emission wavelengths, and the locations of the LED dies; And e) predicting the LED emission wavelength of the LED structures formed on the product wafer using the relationship determined in act d). The method of claim 7, further comprising sorting the LED structures of the product wafer based on the relationship determined in act d). The method of claim 7, wherein the product wafer comprises a substrate, and further comprising measuring a curvature of the substrate prior to forming the product wafer. The method of claim 9, further comprising measuring at least one of the curvature of the substrate and the curvature of the wafer of the product using coherent gradient sensing (CGS). The method of claim 9, wherein the device structures have a size and include determining the at least one test product wafer characteristic at a size equal to or less than one of the device structural dimensions. A method of forming a semiconductor light emitting device (LED), comprising: forming a product wafer comprising an LED structure formed on a substrate by one of a process variable having one or more process variables, wherein the substrate is formed on the substrate Before the LED structure, the substrate has a known initial curvature C ₀ (x, y); after the light-emitting device structure is formed on the product wafer, the curvature C(x, y) of the product wafer is measured, and Determining a curvature change ΔC(x, y)=C(x,y)-C ₀ (x,y); calculating one of the wafers of the product based on the curvature change ΔC(x,y) of the measurement Stress S(x, y); a relationship between the calculated stress S(x, y) and the (x, y) position of the illuminator structures on the product wafer such that the LED structures An emission wavelength λ _{E is} associated with the calculated stress S(x, y); and comparing the emission wavelength λ _E with an emission wavelength variation tolerance, and sorting the LED structures to one or Multiple boxes. The method of claim 12, further comprising forming the LED device using only the LED structures in one or more selected bins. The method of claim 12, further comprising adjusting at least one of the one or more process variables to reduce the structure of the illuminating devices on the product wafer in the illuminating device structure One of the changes in the emission wavelength as a function of the (x, y) position. The method of claim 14, wherein the process variables are selected from the group consisting of process variables including time, temperature, temperature uniformity, gas partial pressure, gas partial pressure uniformity, gas flow rate, and gas flow. Uniformity. The method of claim 14, further comprising: performing a curvature measurement of one or more test product wafers in the same manner as the product wafer; dividing the one or more test product wafers to form an LED Grains; and measuring the emission wavelength of the LED dies to be in the one or more test products The (x,y) position of the LED dies on the wafer is a function of the measured product wafer curvature and one of the emission wavelengths. The method of claim 16, further comprising: forming a new product wafer; measuring the curvature of the new product wafer; and determining the curvature based on the measurement of the new product wafer on the new product wafer The LED emission wavelength of the LED structure. A method of forming a semiconductor light emitting device (LED), comprising: forming a product wafer comprising an LED structure formed on a semiconductor substrate by one of a process variable having one or more process variables; forming the wafer on the product wafer Measuring a curvature uniformity of the product wafer after the LED structure; and adjusting at least one of the one or more process variables to meet at least one of a curvature uniformity requirement and a stress uniformity requirement The difference between the emission wavelength of one of the LED structures and a desired emission wavelength is no more than +/- 2 nm. A method of forming a semiconductor light emitting device (LED), comprising: forming a product wafer comprising an LED structure formed on a substrate by one of a process variable having one or more process variables, wherein the substrate is formed on the substrate Before the structure of the illuminating device, the substrate has a known initial curvature C ₀ (x, y); after forming the illuminating device structure on the product wafer, measuring a curvature C(x, y) of the product wafer, and Determining a curvature uniformity based on C ₀ (x, y) and C (x, y); determining a first number of grains belonging to a first range of curvature uniformity; determining a second range belonging to one of curvature uniformity a second number of dies; and assigning a quality value to the product wafer based on the first number and the second number. The method of claim 19, further comprising arranging the product wafer based on the assigned quality value. The method of claim 19, further comprising: dividing the product wafer to form LED dies associated with the first number and the second number; and the LED crystals associated with the first number The pellets are for a first application, and the LED dies associated with the second number are used for a second application. A method of forming a semiconductor light emitting device, comprising: forming a product wafer comprising a light emitting device structure formed on a substrate by one of a process variable having one or more process variables, wherein the light emitting device structures are formed on the substrate The substrate has a known initial curvature C ₀ (x, y); after forming the light-emitting device structure on the product wafer, measuring a curvature C(x, y) of the product wafer, and determining a curvature Changing ΔC(x, y) = C(x, y) - C ₀ (x, y); based on the calculated curvature C(x, y) and the structure of the illuminators on the wafer of the product ( a relationship between the positions of x, y) such that the emission wavelength λ _E of one of the illuminating device structures is associated with the calculated wafer curvature C(x, y); and the emission wavelength λ _E and an emission wavelength are varied The tolerance comparison is to determine which illuminating structures can be used to form a illuminating device.