US 20100240227 A1
A method of fabricating an integrated circuit includes providing a gate conductor spaced above a semiconductor substrate by a gate dielectric, a pair of dielectric spacers disposed on sidewall surfaces of the gate conductor, and source and drain regions disposed in the substrate on opposite sides of the dielectric spacers, wherein the gate conductor and the source and drain regions comprise dopants; and subjecting at least a portion of the dopants to at least 3 consecutive anneal exposures to activate the dopants, wherein a duration of each exposure is about 200 microseconds to about 5 milliseconds.
1. A method of fabricating an integrated circuit, comprising:
providing a gate conductor spaced above a semiconductor substrate by a gate dielectric, a pair of dielectric spacers disposed on sidewall surfaces of the gate conductor, and source and drain regions disposed in the substrate on opposite sides of the dielectric spacers, wherein the gate conductor and the source and drain regions comprise dopants; and
subjecting at least a portion of the dopants to at least 3 consecutive laser anneal exposures to activate the dopants, wherein a duration of each exposure is about 200 microseconds to about 5 milliseconds; and
wherein each anneal exposure comprises randomizing an exposure pattern of the semiconductor substrate by multiple raster scanning passes in an arc-like fashion in which shifting subsequent passes by a laser beam are shifted with respect to previous passes.
3. The method of
4. The method of
5. The method of
11. The method of
12. The method of
13. The method of
14. The method of
This invention relates to semiconductor fabrication, and particularly to methods of activating dopants using multiple consecutive millisecond-range anneals.
There are ongoing efforts to reduce the dimensions and increase the density of integrated circuit features. Modern integrated circuits (ICs) can include various types of active components such as n-type and p-type field effect transistors (NFETs and PFETs) as well as passive components such as resistors, diodes, electrical fuses, etc. The performance and variability of these components depend on the ability to uniformly activate dopants in semiconductor materials across different microstructures. Due to ongoing efforts to reduce the dimensions and increase the density and complexity of integrated circuit features, there is a continued need to increase the activation of such dopants without inducing excessive diffusion thereof. It is also desirable to maintain a high level of dopant activation across different devices and microstructures.
One specific example of the use of dopants is in the gate conductor and the source and drain regions on opposite sides of the channel of a MOSFET (metal oxide semiconductor field effect transistor) device. As MOSFET devices are scaled down to less than 100 nanometers in gate or channel length, highly doped, shallow source and drain extension regions can be employed to achieve high drive current capability. Currently, shallow source and drain extension regions are formed through the ion implantation of dopants into a semiconductor substrate near its surface on opposite sides of the gate conductor. The dopants are then activated by conducting laser annealing or other millisecond-scale annealing of the implanted extension regions either prior, during, or after a Rapid Thermal Anneal (RTA).
Laser annealing of semiconductors has been widely known in the art for the past several decades. However, up until very recently, it has not been employed in the fabrication of CMOS-based logic and memory ICs due to its large pattern effects, i.e., the sensitivity of laser energy coupling to layout patterns of various microstructures. These large pattern effects can lead to a highly non-uniform heating of various electrical devices and associated microstructures present on the wafer surface. Laser annealing can be characterized by the duration of exposure to its radiation. Pulsed lasers operate in a nanosecond-range regime with exposure durations of tens to hundreds of nanoseconds. At such short anneals, thermal activation of dopants can be inefficient. Consequently, the dopant activation process relies on a phase transition such as melting-recrystallization or solid phase epitaxial (SPE) re-growth of amorphized and doped semiconductors. Due to this reason, nanosecond-scale laser annealing is also referred to as melt laser annealing or pulsed laser annealing. Nanosecond-scale laser annealing has a very large temperature pattern effect because the laser energy absorbed in surface microstructures does not have sufficient time to spread uniformly within the substrate via thermal diffusion. In addition to large pattern effects, its reliance on inducing phase transitions in microstructures produces substantially different levels of dopant activation near exposure edges or in areas of exposure overlap. Nanosecond-scale laser annealing is usually operated in a step-and-repeat mode where a small portion of the wafer surface (typically entire IC rectangular die) is exposed to a pulse at once, followed by a step-and-repeat process to cover the entire wafer surface. This places undesirable areas of exposure overlap or exposure perimeter into the dicing channels that are not electrically usable.
In contrast, millisecond-scale laser annealing has exposure times ranging from several microseconds to tens of milliseconds. In this range, thermal activation of dopants can be efficient, and the concentration of active dopants is proportional to the peak anneal temperature. Continuous wave lasers are employed in this regime. Since the laser beam is shaped in the form of a line, the wafer surface is raster scanned, which means that it is scanned as a pattern of parallel lines or curves. In this case, the exposure time (also referred to as the dwell time) is equal to the characteristic beam width in the scanning direction (often defined at full width at half maximum (FWHM)) divided by the scan speed. The beam length (e.g., about 10 millimeters (mm)) perpendicular to the scanning direction (often defined at full width at 95-99% of the maximum) is usually much smaller than the wafer size (e.g., about 300 mm). As such, adjacent scans (also referred to as exposures) are often applied with some overlap to completely cover the entire wafer surface. In the overlap region, the wafer surface is exposed and annealed twice. Successful application of millisecond-scale laser annealing in IC fabrication depends on whether the overlap region has substantially the same properties as singly annealed regions. In contrast to nanosecond-scale laser annealing, this appears to be true for source and drain activation processes since common dopants undergo little diffusion during laser annealing and their activation depends on the peak anneal temperature in the absence of or after any phase transition processes in the doped material (e.g., after SPE process). Pattern effects can be reduced in millisecond-scale laser annealing by, for example, allowing more time for absorbed laser energy (i.e., heat) to evenly spread in the substrate via thermal diffusion.
In an embodiment, a method of fabricating an integrated circuit comprises: providing a gate conductor spaced above a semiconductor substrate by a gate dielectric, a pair of dielectric spacers disposed on sidewall surfaces of the gate conductor, and source and drain regions disposed in the substrate on opposite sides of the dielectric spacers, wherein the gate conductor and the source and drain regions comprise dopants; and subjecting at least a portion of the dopants to at least 3 consecutive anneal exposures to activate the dopants, wherein a duration of each exposure is about 200 microseconds to about 5 milliseconds.
Additional features and advantages are realized through the techniques of the present invention. Other embodiments and aspects of the invention are described in detail herein and are considered a part of the claimed invention. For a better understanding of the invention with advantages and features, refer to the description and to the drawings.
The subject matter which is regarded as the invention is particularly pointed out and distinctly claimed in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the invention are apparent from the following detailed description taken in conjunction with the accompanying drawings in which:
The detailed description explains the preferred embodiments of the invention, together with advantages and features, by way of example with reference to the drawings.
Turning now to the drawings in greater detail, it will be seen that
Subsequently, source and drain (S/D) regions 300 can be formed in the substrate 50 as described below that include dopants opposite in type from the dopants present in the channel region 200. Dopants can be introduced into the shallow S/D extensions 301 and an upper portion 105 of the gate conductor 100 using, e.g., an ion implantation technique. For only the embodiment depicted in
N-type shallow S/D extensions for NFETs can be formed using a lithography technique to open the pre-selected regions where NFETs are to be formed and to block other regions with a photoresist or hard mask. N-type dopants can then be implanted into exposed regions of the substrate using an implantation energy of about 0.5 kilo electronVolt (keV) to about 10 keV and a dose of about 3e14 to about 1e16 ions/centimeters squared (cm2). Similarly, p-type shallow S/D extensions can be formed by opening pre-selected regions where PFETs are to be formed and implanting p-type dopants using an implantation energy of about 200 eV to about 2 keV and a dose of about 3e14 to about 1e16 ions/cm2. Deep S/D regions for both NFETs and PFETs can be formed in a similar way by adjusting the implantation energy to produce a desired depth. Examples of n-type dopants include but are not limited to arsenic and phosphorus, and examples of p-type dopants include but are not limited to boron and boron difluoride (BF2). It is to be understood that both NFET and PFET devices can be formed in isolated areas of the substrate to form a CMOS (complementary metal-oxide semiconductor) IC. Various known ion implantation enhancement techniques can be employed to beneficially engineer defects in or around implanted layers, shape as-implanted dopant profiles, and alter thermal diffusion at later thermal steps. These techniques include but are not limited to angled, molecular, cluster, pre- and post-amorphization, co-implantation of different species, cold, hot, and plasma implantation processes.
As shown in
Although not shown, precision resistors can also be formed in ICs that provide various analog functions such as impedance matching. These resistors can be formed using either isolated islands of doped semiconductor substrate or doped semiconductor material deposited upon the substrate. The islands can be connected to other elements using conductive interconnects. The resistance value of these devices depends on the geometry of the island and the amount of active dopants in the semiconductor material. The islands can be formed using lithography and etch techniques, and the dopants can be introduced therein either during S/D implantation steps or using a separate block mask with a specialized implant. The dopant dose employed in these islands can be about 1e14 to about 1e16 ions/cm2. It is desirable that the thermal activation of such dopants does not introduce substantial variation of resistor values due to uneven activation across a given IC surface.
While the examples described herein are given using planar transistor geometry, other transistor gate, channel, and source/drain geometrical structures can also be employed. For instance, various multi-gate 3-dimensional structures of the channel and the gate regions are contemplated. Other useful structures can be present in various parts of the transistor regions and other parts of the substrate. For instance, the S/D regions can include stressors, the channel region can include low mobility structures, and/or the gate region can include metallic elements. Some elements of these structures also can be sacrificial and thus would be disposed, replaced, or altered during later steps. The substrate can further include buried insulator regions, isolation regions with or without stressors, deep-trench capacitors, and/or diodes. In addition, the dopants and co-dopants in the S/D regions can be introduced by techniques other than ion implantation such as in-situ doping during epitaxial growth, gas phase doping, or solid source doping.
Referring back to
Subsequent to the first anneal sequence, a second anneal sequence can be conducted that is directed toward higher thermal activation in the doped regions without inducing a significant amount of diffusion. The second anneal sequence can be performed immediately after the first anneal sequence or after some additional processing steps if desired. The strategy of a person of ordinary skill in the art would probably be to use a high-temperature anneal lasting for a short period of time so as to achieve high thermal activation while ensuring the absence of excessive diffusion. As applicable to msec-scale laser anneals, such strategy suggests selecting a high-temperature anneal of about 1300° C. to 1375° C. with a short dwell time of about 50 μsec to about 600 μsec.
It has been unexpectedly discovered that multiple (e.g., greater than 3 exposures or anneals) consecutive msec-range laser anneals of n-type and p-type doped semiconductors, such as silicon and its alloys, to a moderate peak temperature of about 1100° C. to about 1350° C. can yield a substantial increase in the activation of n-type dopant while yielding minimal change in diffusion of the respective p-type dopant. As the number of consecutive laser anneals increases from 2 to 12, the sheet resistance of one-dimensional n-type S/D regions decreases by about 15% to about 20% while showing negligible lateral diffusion of p-type dopant as measured by PFET extension overlap capacitance. By way of example, boron in PFET S/D extensions diffuses faster than arsenic in NFET S/D extensions at a given anneal temperature and anneal time. It has been discovered that incremental dopant diffusion is not induced and the activation of n-type dopants is substantially increased when multiple anneals are employed. It has also been discovered that the effect is stronger at longer dwell times. Accordingly, the dwell or exposure time of the anneal is desirably about 200 μsec to about 5 msec, more specifically about 0.5 msec to about 1.5 msec. Further, it has been discovered that the effect is stronger if the thermal budget in the first anneal sequence is lower. Thermal budget is lower when either the anneal temperature or the anneal time is lower. In addition, it has been discovered that the strongest relative effect occurs when the first anneal sequence includes only msec-range anneals that fully re-crystallize amorphous semiconductor layers. Increasing the number of consecutive anneals from 12 to 24 provides a marginal additional improvement in n-type dopant activation. Therefore, the msec-range anneal is desirably repeated for about 3 to about 12 times.
It has also been discovered that the activation benefit obtained from 12 consecutive anneals at a given moderate peak temperature is equivalent to that obtained through single or double laser exposure anneals performed at a temperature 100° C. to 200° C. higher than the temperature of the 12 consecutive anneals. For instance, in this respect, a 1250° C. anneal conducted 12 times is equivalent to a 1375° C. double anneal. Semiconductor wafer builds can exhibit a slip threshold of about 1325° C. to about 1350° C., a melt threshold of about 1300° C. to about 1350° C. for SiGe semiconductor alloys that are often present in these wafer builds, and gate dielectric degradation threshold of about 1300° C. to about 1350° C. Due to such undesirable effects, single or double msec-scale anneals are desirably kept at or below about 1300° C. to about 1350° C. In contrast, multiple, consecutive msec-scale anneals do not appear to induce any substantial undesirable effects as compared to a double anneal conducted at the same peak temperature. Therefore, multiple, consecutive msec-scale anneals can be used to obtain substantial dopant activation at a moderate anneal temperatures of about 1,150° C. to about 1300° C., thereby opening up the process window with respect to undesirable defect creation at the high temperature range of at or above 1300° C. to about 1350° C.
It has additionally been observed that the multiple, consecutive msec-scale anneals have a substantial advantage over a single anneal conducted at the same peak temperature but for a longer time. Further, it has been discovered that certain structural characteristics of semiconductor substrates exposed to multiple, consecutive msec-scale anneals do not show any degradation with increased number of such anneals while some of these structural characteristics show a consistent degradation with increased anneal time. More specifically, these structural characteristics of the substrate include its flatness, degree of distortion, and structural integrity (absence of cracks, for instance). The first two parameters are related to inducing substantial plastic deformation within the substrate that results in its permanent deformation or distortion. Once the substrate is deformed and/or distorted, subsequent precision lithographic processes yield increased error and variability of certain patterning parameters such as increased pattern overlay error and increased overlay scatter. Accordingly, the lithographic overlay error and its scatter measured at the contact via lithography step, for instance, can serve as a direct and sensitive parameter for assessing degree of wafer distortion induced in prior processing steps, including the step with multiple laser anneals. Alternatively, the wafer warpage or bow can be measured before and after the laser anneal step. The difference in wafer bow or warpage provides a measure of induced wafer distortion. Further, a direct observation of evolving micro and macro slip planes and other crystallographic plane and line defects via various microscopic techniques can also provide a measure of induced wafer distortion. Assessment of these structural parameters yield a conclusion that multiple, consecutive msec-scale anneals do not degrade pattern overlay errors and tolerances in subsequent lithographic steps and do not increase wafer warpage or bow with increasing number of such anneals for certain “prime” semiconductor substrates.
Patterned “prime” substrates or product wafer builds are characterized by a low concentration of certain types of stacking faults and dislocations in its bulk. It has been observed that threading dislocations associated with stress relaxation in crystalline films and/or extrinsic stacking faults bordered by dislocations and associated with precipitation of self-interstitials should be at a low concentration in the upper portion of the substrate of approximately 10 to 20 μm from its top surface to allow for conducting multiple laser anneals without any wafer distortion. The concentration of these defects depends on both starting substrates, microstructures formed on the substrate surface, and processes conducted prior to the laser anneal. In one embodiment, an area density of such defects can be equal to or lower than 105 cm−2 at any plane in the upper 10 to 20 microns of substrate thickness. In contrast, it has been observed that crystal originated particles (COPs) that are often present in the bulk portion of the substrate do not interfere with multiple laser anneals. It has also been observed that these substrates could contain edge defects (e.g. edge slip lines), typically within 3 mm of the wafer edge, that do not substantially grow or multiply when exposed to multiple laser anneals. It has further been discovered that large edge defects (e.g. cracks) result in catastrophic wafer breakage that occurs during the first anneal, hence, yielding little dependence of the wafer breakage on the number of exposures. A preferred prime SOI substrate is a bonded-type SOI substrate with a low concentration of aforementioned defects, as specified above, in the handler wafer (i.e. below the buried oxide). A preferred bulk substrate has a low concentration of threading dislocations as specified above in microstructures formed on its surface.
These multiple laser anneals can also be employed to reduce pattern effects and variations in the resistance value of the precision resistor. The wafer exposure pattern can be beneficially randomized or offset from exposure to exposure in such a way that the beam trajectory does not repeat substantially the same path multiple times. This randomization can be used, for example, by using small stepping increments, by exposing with dissimilar (e.g. orthogonal) trajectories or by using a hybrid exposure pattern where both small increments and dissimilar trajectories are combined. In a first example illustrated in
In a second example (not shown), a similar exposure pattern can be realized via one pass but with small beam steps of about ⅛ of beam length. Multiple scans near the wafer edge can still be needed. In a third example shown in
It is to be understand that while the multiple, consecutive msec-scale anneals are described above in reference to laser anneals, such anneals could also be performed using other millisecond-range anneal tools such as a flash lamp tool or an electron-beam tool.
The following non-limiting example further illustrates the various embodiments described herein.
The dopant activation method described herein was reduced to practice using 45-nm high-performance CMOS technology. A plurality of PFETs and NFETs were built on an SOI substrate. Shallow S/D extensions offset from each gate conductor by a first dielectric spacer were formed utilizing ion implantation of arsenic and BF2 (molecular) for the NFET and PFET devices, respectively. Deep S/D regions were offset from the gate conductor by the first dielectric spacer and a second dielectric spacer were formed by ion implantation of arsenic and boron. The NFET polysilicon gate conductors were additionally implanted with phosphorus ions in order to provide a specified high concentration of n-type dopants near the gate dielectric interface. Precision resistors were formed by implanting boron with a specialized block mask to complete IC fabrication. The foregoing IC fabrication was repeated on different SOI substrates to form several samples. Each sample was subjected to a first anneal sequence that included a furnace anneal and an RTA. These anneals were conducted non-consecutively, i.e., other IC fabrication steps were performed between the anneals. The combined thermal budgets of these anneals enabled full re-crystallization of the amorphous regions in the substrate and creation of a 1 to 3 nm overlap between the S/D extensions and the gate edges for all NFETs and PFETs. Subsequently, each sample was subjected to either 8 or 12 consecutive laser anneals using a dwell time of about 0.8 msec and either four or six repeating passes with double scanning in each pass, respectively. The anneal temperature was different for each sample and was selected conservatively to be within the 1200° C. to 1300° C. range, avoiding any chance of degrading reliability parameters and inducing defects. Additional samples were used as references by subjecting them to a single pass (single double scan) laser anneal at different temperatures. After this annealing sequence, additional IC fabrication steps were performed that did not exceed 480° C.
At the same time, no defects associated with the multiple msec-range anneals were observed in the “prime” SOI substrates used in this example. These substrates are characterized by a low concentration of certain defects at immediately prior to the laser annealing step, as alluded above.
As used herein, the terms “a” and “an” do not denote a limitation of quantity but rather denote the presence of at least one of the referenced items. Moreover, ranges directed to the same component or property are inclusive of the endpoints given for those ranges (e.g., “about 5 nanometers (nm) to about 20 nm,” is inclusive of the endpoints and all intermediate values of the range of about 5 nm to about 20 nm). Reference throughout the specification to “one embodiment”, “another embodiment”, “an embodiment”, and so forth means that a particular element (e.g., feature, structure, and/or characteristic) described in connection with the embodiment is included in at least one embodiment described herein, and might or might not be present in other embodiments. In addition, it is to be understood that the described elements may be combined in any suitable manner in the various embodiments. Unless defined otherwise, technical and scientific terms used herein have the same meaning as is commonly understood by one of skill in the art to which this invention belongs.
While the preferred embodiment to the invention has been described, it will be understood that those skilled in the art, both now and in the future, may make various improvements and enhancements which fall within the scope of the claims which follow. These claims should be construed to maintain the proper protection for the invention first described.