COUPLING OF ACOUSTIC SENSOR FOR CHEMICAL MECHANICAL POLISHING TECHNICAL FIELD [0001] This disclosure relates to in-situ monitoring of chemical mechanical polishing, and in particular to acoustic monitoring. BACKGROUND [0002] An integrated circuit is typically formed on a substrate by the sequential deposition of conductive, semiconductive, or insulative layers on a silicon wafer. One fabrication step involves depositing a filler layer over a non-planar surface and planarizing the filler layer. For certain applications, the filler layer is planarized until the top surface of a patterned layer is exposed. A conductive filler layer, for example, can be deposited on a patterned insulative layer to fill the trenches or holes in the insulative layer. After planarization, the portions of the metallic layer remaining between the raised pattern of the insulative layer form vias, plugs, and lines that provide conductive paths between thin film circuits on the substrate. For other applications, such as oxide polishing, the filler layer is planarized, e.g., by polishingfor a predetermined time period, to leave a portion of the filler layer over the nonplanar surface. In addition, planarization of the substrate surface is usually required for photolithography. [0003] Chemical mechanical polishing (CMP) is one accepted method of planarization. This planarization method typically requires that the substrate be mounted on a carrier or polishing head. The exposed surface of the substrate is typically placed against a rotating polishing pad. The carrier head provides a controllable load on the substrate to push it against the polishing pad. An abrasive polishing slurry is typically supplied to the surface of the polishing pad. [0004] One problem in CMP is determining whether the polishing process is complete, i.e., whether a substrate layer has been planarized to a desired flatness or thickness, or when a desired amount of material has been removed. Variations in the slurry distribution, the polishing pad condition, the relative speed between the polishing pad and the substrate, and the load on the substrate can cause variations in the material removal rate. These variations, as well as variations in the initial thickness of the substrate layer, cause variations in the time needed to reach the polishing endpoint. Therefore, the polishing endpoint usually cannot be determined merely as a function of polishing time. [0005] In some systems, the substrate is monitored in-situ during polishing, e.g., by monitoring the torque required by a motor to rotate the platen or carrier head. Acoustic monitoring of polishing has also been proposed. SUMMARY [0006] In one aspect, a chemical mechanical polishing apparatus includes a platen to support a polishing pad, a carrier head to hold a surface of a substrate against the polishing pad, a motor to generate relative motion between the platen and the carrier head so as to polish an overlying layer on the substrate, an in-situ acoustic monitoring system comprising an acoustic window having a top surface to contact the substrate, and a controller configured to detect a polishing endpoint based on received acoustic signals from the in-situ acoustic monitoring system. [0007] In another aspect, a chemical mechanical polishing apparatus includes a platen, a polishing pad supported on the platen, a carrier head to hold a surface of a substrate against the polishing pad, a motor to generate relative motion between the platen and the carrier head so as to polish an overlying layer on the substrate, an in-situ acoustic monitoring system comprising an acoustic sensor that receives acoustic signals from the surface of the substrate, and a controller configured to detect a polishing endpoint based on received acoustic signals from the in-situ acoustic monitoring system. The acoustic sensor is adhesively attached to a bottom surface of the polishing pad. [0008] Implementations may include one or more of the following features. The sensor may be secured to the polishing pad, e.g., by an adhesive. The acoustic window may have a diameter smaller than the acoustic sensor. The acoustic sensor may be a piezoelectric acoustic sensor. The polishing endpoint may be an exposure of an underlying layer due to the polishing of the substrate. The controller may be configured to, in response to the detection, adjust a pressure of the carrier head or adjust a baseline pressure of a subsequent polishing of a new substrate. [0009] One or more of the following possible advantages may be realized. Signal strength of an acoustic sensor can be increased. Acoustic coupling between the polishing layer and the sensor can be established more reliably. Exposure of an underlying layer can be detected more reliably. Polishing can be halted more reliably, and wafer-to-wafer uniformity can be improved. Polishing parameters can be varied upon detection of planarization, i.e., smoothing of the substrate surface, which can improve uniformity or increase the polishing rate. Polishing can be halted upon detection of planarization or after expiration of a preset time following detection of planarization. This can provide an alternative endpoint technique. [0010] The details of one or more implementations are set forth in the accompanying drawings and the description below. Other aspects, features, and advantages will be apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF THE DRAWINGS [0011] FIG.1 illustrates a schematic cross-sectional view of an example of a polishing apparatus. [0012] FIG.2A illustrates a schematic cross-sectional view of an acoustic monitoring sensor that engages a portion of a polishing pad. [0013] FIG.2B illustrates a schematic cross-sectional view of another implementation of an acoustic monitoring sensor that acoustically transmissive layer. [0014] FIG.2C illustrates a schematic cross-sectional view of another implementation of an acoustic monitoring sensor. [0015] FIG.2D illustrates a schematic cross-sectional view of another implementation of an acoustic monitoring sensor in which an acoustic window is formed in the polishing layer and an acoustically transmissive layer is formed in the backing layer of the polishing pad.. [0016] FIG.3 illustrates a schematic top view of a platen having multiple acoustic monitoring sensor windows. [0017] FIG.4 illustrates a schematic top view of a platen having a planar portion surrounding the acoustic monitoring sensor window. [0018] FIGS.5A-5C illustrate planarization of a surface of a substrate. [0019] FIG.6 illustrates a graph of a sum of spectral power density over a frequency range as a function of time. [0020] Like reference symbols in the various drawings indicate like elements. DETAILED DESCRIPTION [0021] In some semiconductor chip fabrication processes an overlying layer, e.g., metal, silicon oxide or polysilicon, is polished until an underlying layer, e.g., a dielectric, such as silicon oxide, silicon nitride or a high-K dielectric, is exposed. For some applications, when the underlying layer is exposed, the acoustic emissions from the substrate will change. The polishing endpoint can be determined by detecting this change in acoustic signal. However, existing monitoring techniques may not satisfy increasing demands of semiconductor device manufacturers. [0022] The acoustic emissions to be monitored can be caused by energy released when the substrate material undergoes deformation, and the resulting acoustic spectrum is related to the material properties of the substrate. Without being limited to any particular theory, possible sources of this energy, also termed “stress energy”, and its characteristic frequencies include breakage of chemical bonds, characteristic phonon frequencies, slip-stick mechanisms, etc. It may be noted that this stress energy acoustic effect is not the same as noise generated by vibrations induced by friction of the substrate against the polishing pad (which is also sometimes referred to as an acoustic signal), or of noise generated by cracking, chipping, breakage or similar generation of defects on the substrate. The stress energy can be distinguished from other acoustic signals, e.g., from friction of the substrate against the polishing pad or of noise generated by generation of defects on the substrate, through appropriate filtering. For example, the signal from the acoustic sensor can be compared to a signal measured from a test substrate that is known to represent stress energy. [0023] However, a potential problem with acoustic monitoring is transmission of the acoustic signal to the sensor. Some polishing pads have poor transmission of acoustic energy. In addition, poor coupling between the polishing pad and the sensor tends to dampen the acoustic signal. Moreover, establishing a consistent coupling from sensor-to-sensor can be difficult. [0024] Thus, it would be advantageous to have the acoustic sensor in contact with an acoustic “window” with low attenuation of the acoustic signal. In some implementations, a second layer of transmissive material is added to the in-situ acoustic monitoring system to further increase the acoustic signal coupling to the acoustic sensor. [0025] Adhering the acoustic sensor to the coupling window, e.g., with an adhesive, can reduce noise in the acoustic signal associated with movement of the acoustic sensor within the housing. The adhesive can provide superior coupling of the sensor to the polishing pad and provide more reliable acoustic attenuation on a sensor-to-sensor basis. [0026] Any of these features could be used independent of the other features. [0027] FIG.1 illustrates an example of a polishing apparatus 100. The polishing apparatus 100 includes a rotatable disk-shaped platen 120 on which a polishing pad 110 is situated. The polishing pad 110 can be a two-layer polishing pad with an outer polishing layer 112 and a softer backing layer 114. The platen is operable to rotate about an axis 125. For example, a motor 121, e.g., a DC induction motor, can turn a drive shaft 124 to rotate the platen 120. [0028] The polishing apparatus 100 can include a port 130 to dispense polishing liquid 132, such as abrasive slurry, onto the polishing pad 110 to the pad. The polishing apparatus can also include a polishing pad conditioner to abrade the polishing pad 110 to maintain the polishing pad 110 in a consistent abrasive state. [0029] The polishing apparatus 100 includes at least one carrier head 140. The carrier head 140 is operable to hold a substrate 10 against the polishing pad 110. Each carrier head 140 can have independent control of the polishing parameters, for example pressure, associated with each respective substrate. [0030] The carrier head 140 can include a retaining ring 142 to retain the substrate 10 below a flexible membrane 144. The carrier head 140 also includes one or more independently controllable pressurizable chambers defined by the membrane, e.g., three chambers 146a-146c, which can apply independently controllable pressurizes to associated zones on the flexible membrane 144 and thus on the substrate 10 (see FIG.1). Although only three chambers are illustrated in FIG.1 for ease of illustration, there could be one or two chambers, or four or more chambers, e.g., five chambers. [0031] The carrier head 140 is suspended from a support structure 150, e.g., a carousel or track, and is connected by a drive shaft 152 to a carrier head rotation motor 154, e.g., a DC induction motor, so that the carrier head can rotate about an axis 155. Optionally each carrier head 140 can oscillate laterally, e.g., on sliders on the carousel 150, or by rotational oscillation of the carousel itself, or by sliding along the track. In typical operation, the platen is rotated about its central axis 125, and each carrier head is rotated about its central axis 155 and translated laterally across the top surface of the polishing pad. [0032] A controller 190, such as a programmable computer, is connected to the motors 121, 154 to control the rotation rate of the platen 120 and carrier head 140. For example, each motor can include an encoder that measures the rotation rate of the associated drive shaft. A feedback control circuit, which could be in the motor itself, part of the controller, or a separate circuit, receives the measured rotation rate from the encoder and adjusts the current supplied to the motor to ensure that the rotation rate of the drive shaft matches at a rotation rate received from the controller. [0033] The polishing apparatus 100 includes at least one in-situ acoustic monitoring system 160. The in-situ acoustic monitoring system 160 includes one or more acoustic signal sensors 162. Each acoustic signal sensor can be installed at one or more locations on the upper platen 120. In particular, the in-situ acoustic monitoring system can be configured to detect acoustic emissions caused by stress energy when the material of the substrate 10 undergoes deformation. [0034] A position sensor, e.g., an optical interrupter connected to the rim of the platen or a rotary encoder, can be used to sense the angular position of the platen 120. This permits only portions of the signal measured when the sensor 162 is in proximity to the substrate, e.g., when the sensor 162 is below the carrier head or substrate, to be used in endpoint detection. [0035] In the implementation shown in FIG.1, the acoustic monitoring system 160 includes an acoustic sensor 162 positioned supported by the platen 120 to receive acoustic signals through the polishing pad 110 from the substrate 10. The acoustic sensor 162 can be partially or entirely in a recess 164 in the top surface of the platen 120. In some implementations, a top surface of the acoustic sensor 162 is coplanar with the top surface of the platen 120. [0036] The portion of the polishing pad directly above the acoustic sensor 162 can include an acoustic window 119. The acoustic window 119 can be narrower than the acoustic sensor 162, e.g., as shown in FIG.2A, or the two can be of substantially equal width (e.g., within 10%), e.g., as shown in FIG.2C. Where the acoustic window 119 is narrower than the acoustic sensor 119, the sensor can also abut the bottom of the polishing layer 112. [0037] The acoustic sensor 162 is a contact acoustic sensor having a surface connected to (e.g., in direct contact with or having just an adhesive layer) a portion of the polishing layer 112 and/or the acoustic window 119. For example, the acoustic sensor 162 can be an electromagnetic acoustic transducer or piezoelectric acoustic transducer. A piezoelectric sensor can include a rigid contact plate, e.g., of stainless steel or the like, which is placed into contact with the body to be monitored, and a piezoelectric assembly, e.g., a piezoelectric layer sandwiched between two electrodes, on the backside of the contact plate. [0038] In some implementations, the acoustic sensor 162 is positioned within a recess 169 in a housing 163. An optional spring 165 can be arranged between the housing 163 and a support 167 provides pressure against the housing 163. The pressure on the housing 163 presses the acoustic sensor 162 into contact with a portion of the polishing pad 110. Alternatively, the spring 165 can press directly against the acoustic sensor 162, e.g., if a housing is not used. In some implementations, the spring 165 is a long travel spring 165 supplying similar pressure as the strong spring 165 over larger compression ranges. [0039] The acoustic sensor 162 can be connected by circuitry 168 to a power supply and/or other signal processing electronics 166 through a rotary coupling, e.g., a mercury slip ring. [0040] In some implementations, the in-situ acoustic monitoring system 160 is a passive acoustic monitoring system. In this case, signals are monitored by the acoustic sensor 162 without generating signals from an acoustic signal generator (or the acoustic signal generator can be omitted entirely from the system). The passive acoustic signals monitored by the acoustic sensor 162 can be in 50 kHz to 1 MHz range, e.g., 200 to 400 kHz, or 200 Khz to 1 MHz. For example, for monitoring of polishing of inter-layer dielectric (ILD) in a shallow trench isolation (STI), a frequency range of 225 kHz to 350kHz can be monitored. [0041] The signal from the sensor 162 can be amplified by a built-in internal amplifier. In some implementations, the amplification gain is between 40 and 60 dB (e.g., 50 dB). The signal from the acoustic sensor 162 can then be further amplified and filtered if necessary, and digitized through an A/D port to a high speed data acquisition board, e.g., in the electronics 166. Data from the acoustic sensor 162 can be recorded at a similar range as that of the generator 163 or at a different, e.g., higher, range, e.g., from 1 to 10 Mhz, e.g., 1-3 MHz or 6-8 Mz. In implementations in which the acoustic sensor 162 is a passive acoustic sensor, a frequency range from 100 kHz to 2 MHz can be monitored, such as 500 kHz to 1 MHz (e.g., 750 kHz). [0042] If positioned in the platen 120, the acoustic sensor 162, can be located at the center of the platen 120, e.g., at the axis of rotation 125, at the edge of the platen 120, or at a midpoint (e.g., 5 inches from the axis of rotation for a 20 inch diameter platen). [0043] Referring now to FIG.2A, further details of the acoustic monitoring system 160 are shown. The acoustic sensor 162 can be held within a recess 169 in a top surface of the housing 163. The housing 163 can assist in proper positioning of the sensor 162. The housing 163 is composed of a rigid, durable material sufficient to protect the acoustic sensor 162 from damage. However, in some implementations, e.g., as shown FIGS.2C and 2D, a housing is not necessary, e.g., the sensor 162 can simply fit between and secured to by the side walls of the recess 169. The various implementations described as using a housing can omit the housing. [0044] In some implementations, e.g., as shown in FIG.2B, the housing 163 extends through the backing layer 114, and in some implementations, e.g., as shown in FIG.2A, the housing 163 extends through a portion of the polishing layer 112. However, in some implementations the housing 163 fits entirely within the recess 164 in the platen 120, e.g., if the top surface of the sensor 163 is coplanar with the top surface of the platen 120 and contact the bottom surface of the polishing pad 110. [0045] In some implementations, the housing 163 material is acoustically dampening to reduce noise received by the acoustic sensor 162 from surfaces in contact with the housing 163, such as the backing layer 114 or polishing layer 112 through which the housing 163 extends. The housing 163 can be composed of a metal, e.g., aluminum or stainless steel, or a polymer material, e.g., polycarbonate, polyvinyl chloride (PVC), or polymethyl-methacrylate (PMMA). [0046] Assuming a spring is used, one end of the spring 165 contacts the housing 163 on a surface opposing the acoustic sensor 162. In some implementations, the other end of the spring 165 is in contact with a support 167 that rests on the platen 120. Such a support can provide a stable base for the forces generated by compression of the spring 165. In some implementations, the other end of the spring 165 is in contact with the bottom surface of the recess 164, i.e., in direct contact with the platen. The spring 165 presses the housing 163 toward a polishing surface 112a of the polishing layer 112, which urges the acoustic sensor 162 into contact with the bottom surface of the polishing layer 112. This can improve the acoustic coupling between the polishing layer and the sensor. However, the various implementations described as using a spring can omit the spring, e.g., assuming the sensor is adhesively attached to the bottom of the acoustic window 119 and/or polishing pad 110. [0047] In some implementations, a support 167 is arranged beneath the spring 165 which provides a stationary block which the spring 165 can push against opposite the housing 163. The support 167 can be any material sufficient to rigidly support the spring 165 and housing 163 without movement or compressive buckling. [0048] In addition to or instead of the spring, the acoustic sensor 162 can be secured to a portion of the polishing layer 112 (and/or to an acoustic window 119 described below) by an adhesive layer 170. The adhesive layer 170 increases the contact area between the acoustic sensor 162 and the polishing layer 112 and/or acoustic window 119, reduces undesirable motion in the acoustic sensor 162 during polishing operations, and can reduce the presence of gas pockets between the acoustic sensor 162 and the polishing layer 112 and/or acoustic window 119 thereby improving the coupling to the sensor, thus reducing noise in the acoustic signal received by the acoustic sensor 162. The adhesive layer 170 can be a glue applied between the acoustic sensor 162 and the polishing layer 112 and/or acoustic window 119, or an adhesive strip (e.g., tape). For example, the adhesive layer 170 can be a cyanocrylate, a pressure sensitive adhesive, a hot melt adhesive, etc.. [0049] Returning to FIG.2A, the polishing layer 112 includes an acoustic window 119 arranged above the adhesive layer 170 and the acoustic sensor 162. However, in some implementations, the acoustic sensor 162 contacts the acoustic window 119 directly. [0050] In implementations with an acoustic window, the acoustic window 119 is formed of a different material than the polishing layer 112. The material of the acoustic window has sufficient acoustic transmission characteristics, e.g., an acoustic impedance of between 1 and 4 MRayl and an acoustic attenuation coefficient lower than 2 (e.g., lower than 1, lower than 0.5). to provide a signal satisfactory for acoustic monitoring. [0051] The acoustic impedance of a material is a measure of the opposition that a material presents to the acoustic flow resulting from an acoustic pressure applied to the material. The acoustic attenuation coefficient quantifies how transmitted acoustic amplitude decreases as a function of frequency for a specific material. Without wishing to be bound by theory, the specific acoustic impedance of the acoustic window 119 (AI _window ) coupling the liquid 132 and polishing surface 112a to the acoustic signal sensor 162, the acoustic window 119 specific acoustic impedance can be beneficially within the range ~ 3െ 6. [0052] In particular, the window 119 can have lower acoustic attenuation than the surrounding polishing layer 112. This permits the polishing layer 112 to be composed of a wider range of materials to meet the needs of the CMP operation. The window can be composed of a non-porous material, e.g., a solid body. In contrast, the polishing layer 112 can be porous, e.g., be microporous, such as a polymer matrix in which hollow plastic microspheres are embedded. [0053] The acoustic window 119 extends through the polishing layer 112 such that one surface, e.g., an upper surface, is coplanar with the polishing surface 112a of the polishing layer 112. The opposing surface, e.g., a bottom surface, can be coplanar with a lower surface 112b of the polishing layer 112. In some implementations, an indentation 118 is formed in the lower surface 112b opposing the polishing surface 112a. The portion of the polishing layer 112 that includes the indentation 118 forms a thin portion of the polishing layer 112 having a thickness that is less than the remaining polishing layer 112, and the acoustic window 119 is located in the thin portion. [0054] The acoustic window 119 can be composed of a non-porous material. In general, non-porous materials transmit acoustic signals with reduced noise and dispersion compared to porous materials. The acoustic window 119 material can have a compressibility within a range of the compressibility of the surrounding polishing layer 112 material that reduces the effect of the acoustic window 119 on the polishing characteristics of the polishing surface on the substrate. In some implementations, the acoustic window 119 compressibility is within 10% of the polishing layer 112 compressibility (e.g., within 8%, within 5%, within 3%). In some implementations, the acoustic window 119 is opaque to light, e.g., visible light. The acoustic window 119 can be composed of one or more of polyurethane, polyacrylate, polyethylene, or other polymers with low acoustic impedance and low acoustic attenuation.. [0055] Referring to FIG.2C, the acoustic window 119 is shown extending through the total thickness of the polishing layer 112 such that the lower surface 112b is planar. The sensor 162 extends through an aperture 114a in the backing layer 114 to contact the underside of the window 119. [0056] In some implementations, the acoustic monitoring system 160 includes an acoustically transmissive layer 172 in contact with the adhesive layer 170. The transmissive layer 172 is an index-matching material which provides increased acoustic signal coupling between the elements in contact with the transmissive layer 172. The transmissive layer 172 can be arranged between the acoustic window 119 and the adhesive layer 170, or between the adhesive layer 170 and the acoustic sensor 162, as shown in FIG.2B. In some implementations, the acoustic monitoring system 160 includes the adhesive layer 170, the transmissive layer 172, or both. For example, the transmissive layer 172 can be a layer of Aqualink™, Rexolite, or Aqualene™. In some implementations, the transmissive layer 172 has an acoustic attenuation that is within 20%, e.g., 10%, of the acoustic attenuation of the acoustic window 119. The acoustically transmissive layer 172 can have an acoustic attenuation less than the acoustic attenuation of the surrounding backing layer 114. [0057] The acoustically transmissive layer 172 can be selected to have a compressibility similar to the compressibility of the backing layer 114, e.g., within 20%, e.g., within 10%, of the compressibility of the surrounding backing layer 114. [0058] FIG.2D is an implementation in which the acoustic window 119 extends through the thickness of the polishing layer 112 and the transmissive layer 172 extends through the thickness of the backing layer 114. However, the transmissive layer 172 could be thinner than the backing layer 114. In this case, the sensor 162 could project above the top surface of the platen 120 to engage the transmissive layer 172. [0059] Additionally, the acoustic signal sensor 162 is shown having dimensions sufficient to contact both the transmissive layer 172 and the opposing surface of the recess 164. In such implementations, the recess 164 provides the support for the acoustic signal sensor 162 while the pressure of the polishing operation brings the acoustic signal sensor 162 into contact with the transmissive layer 172. Arranged between the transmissive layer 172 and the acoustic window 119 is an adhesive layer 170, as those described herein. In some implementations, additional adhesive affixes the contact surface between the acoustic signal sensor 162 and the transmissive layer 172. [0060] In some implementations, the acoustic monitoring system 160 includes an active acoustic monitoring system. Such implementations include an acoustic signal generator and an acoustic sensor, such as acoustic sensor 162. [0061] The acoustic signal generator generates (i.e., emits) acoustic signals from a side of the substrate closer to the polishing pad 110. The acoustic signal generator can be connected by circuitry 168 to a power supply and/or other signal processing electronics 166 through a rotary coupling, e.g., a mercury slip ring. The signal processing electronics 166 can be connected in turn to the controller 190, which can be additionally configured to control the magnitude or frequency of the acoustic energy transmitted by the generator, e.g., by variably increasing or decreasing the current supply to the generator. The acoustic signal generator 163 and acoustic sensor 162 can be coupled to one another, though this is not required. The sensor 162 and the generator can be decoupled and physically separated from one another. For the generator, commercially available acoustic signal generators can be used. The generator can be attached to platen 120 and held in place, e.g., with a clamp or by threaded connection to the platen 120. [0062] As depicted in FIG.3, in some implementations a plurality of acoustic signal sensors 1623 can be installed in the platen 120, each acoustic sensor 162 being associated with an acoustic window 119. Each sensor 162 can be configured in the manner described for any of FIGS.1 and 2A-2B. The signals from the sensors 162 can be used by the controller 190 to compute the positional distribution of acoustic emission events occurring on the substrate 10 during polishing. In some implementations, the plurality of sensors 162 can be positioned at different angular positions around the axis of rotation of the platen 120, but at the same radial distance from the axis of rotation. In some implementations, such as the implementation of FIG.3, the plurality of sensors 162 are positioned at different radial distances from the axis of rotation of the platen 120, but at the same angular position. In some implementations, the plurality of sensors 162 are be positioned at different angular positions around and different radial distances from the axis of rotation of the platen 120. [0063] In some implementations, the acoustic window 119 is surrounded by a smooth portion 174 of the polishing layer 112. The smooth portion 174 lacks grooves 116 and is coplanar with the upper surface of the acoustic window 119. Implementations including a smooth portion 174 surrounding the acoustic window 119 can reduce noise associated with the substrate 10 interacting with the grooves 116 of a polishing layer 112 during a polishing operation. [0064] A substrate 10 is formed by the sequential deposition of conductive, semiconductive, or insulative layers on a silicon wafer. A filler layer is deposited over a non- planar surface and planarized such that the filler and non-planar surface, such as a patterned layer, have a common coplanar surface and/or the non-planar surface is exposed. In some implementations, the in-situ acoustic monitoring system 160 detects transitions between layers, or topography information related to one or more layers of the substrate 10. This provides information to be used between process steps. For example, a substrate 10 including a filler layer can have non-uniform surface roughness, e.g., topography, from the deposition process. Detecting when the topography has been planarized allows the system to modify one or more process conditions based on the transition. For example, the apparatus 100 can stop a high carrier head 140 pressure step once the filler layer surface has been planarized. [0065] FIGS.5A-5C depict intermediary layer transitions present in a planarization process for a substrate 500. FIG.6 depicts an exemplary acoustic signal 600 comparing the summed power spectral density (PSD) across a frequency range on the y-axis against time, in seconds (s). The acoustic signal 600 has distinct regions, such as first region 602, second region 604, and third region 606. In some implementations, the regions 602, 604, and 606 correspond to layer transitions in the substrate 500, such as those depicted in FIGS.5A-5C. [0066] FIG.5A shows an exemplary substrate 10 prior to polishing. The substrate 10 include a wafer 502, e.g., a silicon wafer, a patterned layer 504, and a filler layer 508. Prior to a planarization step, the filler layer 508 is non-planar and includes topography 509. The topography 509 can result from deposition of the filler layer 508 over the patterned layer 504, and has dimensions on order of the feature size, e.g., the metal line width. [0067] During operation the carrier head holds the substrate 10 and relative motion is generated between the polishing layer 112 and the substrate 10. The acoustic signal sensor receives an acoustic signal, such as acoustic signal 600, based upon the contact of the polishing surface 112a and the outer-most layer of the substrate 10. In FIG.5A, at the initiation of polishing, the topography 509 and the polishing layer 112 are in contact. [0068] Without wishing to be bound by theory, the acoustic signal 600 changes based upon the changing contact surface of the filler layer 508 material and the polishing layer 112 material. In particular, initially the uneven topography may create a significant acoustic signal. However, as polishing progresses and the topography 509 of the filler layer 508 is planarized, the interface between the polishing surface 110 and the substrate 10 become smoother, and the acoustic signal may decrease. The polishing of the topography 509 may corresponds to a first region 602 of the signal 600 in FIG.6. [0069] Again without wishing to be bound by theory, a layer transition occurs when the topography 509 has been removed by the apparatus 100. As shown in FIG.5B, the surface of the remaining filler layer 508 is substantially planar. The polishing of the planar surface may correspond to a second region 604 of the acoustic signal 600. In the second region 604 of the signal 600 the acoustic signal 600 is substantially constant (albeit subject to noise). [0070] Still without wishing to be bound by theory, the second region 604 continues in time until the filler layer 508 extending above the patterned layer 504 has been removed. As shown in FIG.5C, the patterned layer 504 is composed of a different material than the filler layer 508 and interacts with the polishing layer 112 surface and materials differently, thereby creating a third region 606 of the acoustic signal 600. In addition, continued polishing can create dishing, and this topology may again increase the acoustic signal. The third region 606 is not constant, e.g., can be increasing or decreasing.. [0071] In some implementations, the differentiation, e.g., detection of the layer transitions, between the regions 602, 604, and 606 can be accomplished by the acoustic monitoring system 160 and/or the controller 190 of the apparatus 100. The detection can be accomplished through various calculations known to the field for detection of slope change, but can include calculation of one or more differential, rolling average, windowing, or box logic algorithms. [0072] In additional implementations, the acoustic signal 600 can be processed using additional steps prior to application of a slope-change detection algorithm. For example, the acoustic signal 600 can be subjected to one or more filters, e.g., a bandpass filter, and/or one or more transformations, e.g., a fast Fourier transformation. For example, the bandpass filter can be used to isolate preferred frequencies of the acoustic signal 600 before processing, such as frequencies in a range from 50 to 500 kHz, or in a range from 200 to 700 kHz. [0073] In some implementations, the apparatus 100 modifies one or more polishing parameter responsive to differentiating the regions 602, 604, and 606. For example, during the first region 602 in which the topography 509 is being removed, the apparatus 100 can dispense first abrasive polishing liquid 132 for rapid removal of the topography 509. Once the transition from the first region 602 to the second region 604 is detected, a different polishing liquid 132 having a lower polishing rate or a lower selectivity can be dispensed to the pad 110. [0074] Alternatively or in addition, once the transition from the second region 604 to the third region 606 is detected, the pressure applied by the carrier head 140 can be reduced. This can reduce the danger of dishing or erosion of the filler layer 508. [0075] Turning now to the signal from the sensor 162 of any of the prior implementations, the signal, e.g., after amplification, preliminary filtering and digitization, can be subject to data processing, e.g., in the controller 190, for either endpoint detection or feedback or feedforward control. [0076] In some implementations, the controller 190 is configured to monitor acoustic loss. For example, the received signal strength is compared to the emitted signal strength to generate a normalized signal, and the normalized can be monitored over time to detect changes. Such changes can indicate a polishing endpoint, e.g., if the signal crosses a threshold value. [0077] In some implementations, a frequency analysis of the signal is performed. For example, frequency domain analysis can be used to determine changes in the relative power of spectral frequencies, and to determine when a film transition has occurred at a particular radius. Information about time of transition by radius can be used to trigger endpoint. As another example, a Fast Fourier Transform (FFT) can be performed on the signal to generate a frequency spectrum. A particular frequency band can be monitored, and if the intensity in the frequency band crosses a threshold value, this can indicate exposure of an underlying layer, which can be used to trigger endpoint. Alternatively, if a location (e.g., wavelength) or bandwidth of a local maxima or minima in a selected frequency range crosses a threshold value, this can indicate exposure of an underlying layer, which can be used to trigger endpoint. For example, for monitoring of polishing of inter-layer dielectric (ILD) in a shallow trench isolation (STI), a frequency range of 225 kHz to 350kHz can be monitored. [0078] As another example, a wavelet packet transform (WPT) can be performed on the signal to decompose the signal into a low-frequency component and a high frequency component. The decomposition can be iterated if necessary to break the signal into smaller components. The intensity of one of the frequency components can be monitored, and if the intensity in the component crosses a threshold value, this can indicate exposure of an underlying layer, which can be used to trigger endpoint. [0079] Assuming the positions of the sensors 162 relative to the substrate 10 are known, e.g., using a motor encoder signal or an optical interrupter attached to the platen 120, the positions of the acoustic events on the substrate can be calculated, e.g., the radial distance of the event from the center of the substrate can be calculated. Determination of the position of a sensor relative to the substrate is discussed in U.S. Patent No.6,159,073 and in U.S. Patent No 6,296,548, incorporated by reference. [0080] Various process-meaningful acoustic events include micro-scratches, film transition break through, and film clearing. Various methods can be used to analyze the acoustic emission signal from the waveguide. Fourier transformation and other frequency analysis methods can be used to determine the peak frequencies occurring during polishing. Experimentally determined thresholds and monitoring within defined frequency ranges are used to identify expected and unexpected changes during polishing. Examples of expected changes include the sudden appearance of a peak frequency during transitions in film hardness. Examples of unexpected changes include problems with the consumable set (such as pad glazing or other process-drift-inducing machine health problems). [0081] In operation, as a device substrate 10 is being polished at the polishing station 100, an acoustic signal is collected from the in-situ acoustic monitoring system 160. The signal is monitored to detect exposure of the underlying layer of the substrate 10. For example, a specific frequency range can be monitored, and the intensity can be monitored and compared to an experimentally determined threshold value. [0082] Detection of the polishing endpoint triggers halting of the polishing, although polishing can continue for a predetermined amount of time after endpoint trigger. Alternatively or in addition, the data collected and/or the endpoint detection time can be fed forward to control processing of the substrate in a subsequent processing operation, e.g., polishing at a subsequent station, or can be fed back to control processing of a subsequent substrate at the same polishing station. For example, detection of the polishing endpoint can trigger modification to the current pressures of the polishing head. As another example, detection of the polishing endpoint can trigger modification to the baseline pressures of the subsequent polishing of a new substrate. [0083] Implementations and all of the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them. Implementations described herein can be implemented as one or more non-transitory computer program products, i.e., one or more computer programs tangibly embodied in a machine readable storage device, for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple processors or computers. [0084] A computer program (also known as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network. [0085] The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit). [0086] The term “data processing apparatus” encompasses all apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, or multiple processors or computers. The apparatus can include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them. Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. [0087] Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto optical disks; and CD ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry. [0088] The above described polishing apparatus and methods can be applied in a variety of polishing systems. Either the polishing pad, or the carrier head, or both can move to provide relative motion between the polishing surface and the wafer. For example, the platen may orbit rather than rotate. The polishing pad can be a circular (or some other shape) pad secured to the platen. Some aspects of the endpoint detection system may be applicable to linear polishing systems (e.g., where the polishing pad is a continuous or a reel-to-reel belt that moves linearly). The polishing layer can be a standard (for example, polyurethane with or without fillers) polishing material, a soft material, or a fixed-abrasive material. Terms of relative positioning are used; it should be understood that the polishing surface and wafer can be held in a vertical orientation or some other orientations. [0089] While this specification contains many specifics, these should not be construed as limitations on the scope of what may be claimed, but rather as descriptions of features that may be specific to particular embodiments of particular inventions. In some implementations, the method could be applied to other combinations of overlying and underlying materials, and to signals from other sorts of in-situ monitoring systems, e.g., optical monitoring or eddy current monitoring systems.