HIGH FREQUENCY LASER BASED ACOUSTIC MICROSCOPY USING A CW GENERATION SOURCE

Transcription

1 HIGH FREQUENCY LASER BASED ACOUSTIC MICROSCOPY USING A CW GENERATION SOURCE T.W. Murray, O. Balogun, and N. Pratt Departent of Aerospace and Mechanical Engineering, Boston University, Boston MA 0225 Abstract: A laser ultrasonic syste has been developed which uses a odulated CW laser source for ultrasonic wave generation. The ajority of high frequency laser ultrasonic systes use picosecond pulsed laser sources to generate ultrasound in the hundreds of MHz to the low GHz range. High frequency SAW generation and high-resolution acoustic icroscopy require that the generation laser spot be tightly focused on the saple surface. This places a severe restriction on the aount of energy that can be deposited in the saple without causing surface ablation and ultiately liits the sensitivity of the syste. In this work, we explore the use of an aplified, electro-absorption odulated diode laser source for high frequency ultrasound generation. The laser can be aplitude odulated up to 2 GHz and the peak power is approxiately 2 Watts. While LBU systes using odulated CW generation can have extreely high SNR when lock-in detection systes are eployed, they have disadvantages when copared to pulsed systes in ters of signal interpretation when ultiple acoustic odes are present or when signals are reflected fro the boundaries of the target aterial leading to coplex acoustic interference patterns. In order to avoid these difficulties the source frequency is scanned over the bandwidth of interest and the transient response of the specien reconstructed fro the frequency doain data. Experiental results are presented for data obtained on thin fils and plates in the 0.KHz- 200MHz range. The acoustic waves are detected using a stabilized Michelson interferoeter fed into an RF lock-in aplifier. The reduced bandwidth of the syste afforded by lock-in detection allows for a substantial iproveent in signal-to-noise ratio over systes using pulsed laser generation for the sae surface teperature rise. The experiental results are copared to theory and a detailed analysis of the SNR of the syste given. Introduction: Conventional laser ultrasonic techniques typically use pulsed laser sources for the generation of acoustic waves [-3]. This is due in part to the fact that odulated laser sources with enough power to generate acoustic waves with sufficient aplitude for detection with an optical probe have not been readily available. Recent developents in laser technology, geared priarily for the teleco sector, have resulted in electro-absorption odulated DFB diode laser sources that can be aplitude odulated at frequencies approaching 40GHz. In addition, erbiu doped fiber aplifiers are available to aplify the output of these lasers. It is noted that these coponents are relatively inexpensive, especially when copared to fetosecond and picosecond solid-state laser systes. High power laser sources that can be odulated at GHz frequencies provide an attractive alternative to pulsed laser sources for laser based acoustic icroscopy.

2 .00 pulsed response CW response Relative Aplitude Tie (us) Figure. Acoustic response of an aluinu half-space showing SAWs generated using CW and pulsed laser sources. The laser sources produce the sae axiu surface teperature in the saple. The reasons for exploring photoacoustic icroscopy using a CW laser source are twofold. First, it is expected that the signal-to-noise ratio of a CW syste will be significantly iproved over that of a pulsed syste in a nuber of cases. Next, it is possible to resonantly excite sall scale structures such as coatings, ebranes, beas used in MEMs applications using a CW laser source to evaluate their echanical properties. For theroelastic generation of acoustic waves, there exists soe teperature T ax (typically taken as the elting point) that the saple surface is kept below in order to avoid daage or ablation. For a given laser pulse shape, this liits the axiu allowable absorbed power density at the surface. As an exaple, we copare laser generation of ultrasound in aluiniu with a 5ns pulsed laser source and a 60 MHz CW laser source. The laser spot size is taken as 3µ. It is found that, for the sae absorbed power density in each case, the CW laser heats the aterial to a teperature of approxiately 3.0 ties higher than the pulsed laser. This is due to the fact that heat builds up in the saple between cycles until the saple reaches steady state. Now, the CW laser power is scaled down by a factor of 3 such that both of the laser sources produce equivalent surface heating. The scaled pulse shapes are then convolved with the ipulse response of an aluinu sei-infinite half space (with the source and receiver slightly offset on the saple surface) to find the acoustic response of the saple. The resulting signals are shown in Figure. As is evident in the pulsed laser case, the laser source produces a strong surface acoustic wave (SAW). For laser powers that produce equivalent surface heating, the SAW displaceent peak-to-peak aplitude is a factor of about 3.0 higher than that of CW generation. However, the bandwidth of the CW signal can be substantially reduced through detection with an RF lock-in aplifier. Using a sufficiently long integration tie, the bandwidth can be reduced by ore than six orders of agnitude for the narrowband case over the broadband case, resulting in a SNR increase of ore than three orders of agnitude for this particular exaple. SNR is an iportant issue in laser based systes, which have lower sensitivity than conventional contact transducers [], and this type of SNR increase could open up the possibility of using these non-contact systes for a uch wider range of inspection applications.

3 Note that the iproved SNR is strongly dependent on the spot size and theral conductivity of the specien. For laser ultrasonic icroscopy, though, sall spot sizes are required to produce localized sources and to ake easureents over short source to receiver distances. It is also noted that the above analysis assues that a high power CW laser source, capable of operating just below the ablation threshold of the saple, is available. Using diffraction liited spot sizes in the µ range, peak power densities in the 00 s of MW/c 2 can be achieved with CW lasers operating in the standard -20W range. One of the disadvantages of CW generation is that the signals can be difficult to interpret, especially when ultiple acoustic arrivals are present. In addition, there is the possibility of setting up coplex interference patterns on the saple surface due to reflections fro saple boundaries. In this work, the CW odulation frequency is scanned over the range of interest and a tie doain response, siilar to that observed when generating with a pulsed laser source, is reconstructed fro the easured frequency doain data. This approach allows for individual arrivals or odes to be easily identified in the data and for tie doain gating of signals for data analysis. We deonstrate non-contact detection of acoustic waves generated with a odulated diode laser source at frequencies up to 200 MHz. Several researchers have deonstrated the feasibility of using odulated CW sources for acoustic wave generation in the khz to low MHz frequency range. Pierce et al.[4], for exaple, deonstrated that it was possible to detect the acoustic signals generated using a diode laser source odulated using a pseudorando binary sequence. The signals were detected using a surface bonded optical fiber and the output was correlated with the sequence to obtain the tie doain displaceent.

4 Figure 2. Scheatic of experiental setup. Magnitude (f) Frequency (MHz) Figure 3. Magnitude of acoustic signals detected as a function of frequency. Results: The experiental setup is illustrated in Figure 2. The syste has three optical paths that lead to the saple surface through the sae 20x objective (NA = 0.4). The first path leads to a CCD caera and allows for optical iaging of the saple surface as well as saple alignent. In the second path, the detection laser light enters the icroscope though a single ode optical fiber, is colliated, and directed to the saple surface. Upon reflection fro the saple the light is returned to a stabilized Michelson interferoeter where the acoustic signal of interest is detected. The detection laser is a 200 W frequency doubled Nd:YAG (λ=532n). The generation laser is colliated, directed to a irror on a gibal ount, sent though a relay lens syste, and directed to the specien. The gibal ount allows for precise control of the generation point within the field of view of the icroscope. The total field of view of the icroscope is 320 x 280 µ. An electroabsorption odulated DFB diode laser with a W fiber aplifier is used for ultrasound generation. The average generation power incident on the specien was approxiately 400W. The generation laser is odulated using a signal generator and has a axiu odulation frequency of 5 GHz. The output signal fro the photodetector is sent to an RF lock-in aplifier with a axiu detection frequency of 200 MHz. The signal generator and lock-in aplifier are controlled using a Labview progra that allows for the odulation frequency to be scanned over the region of interest. The laser source was scanned over the -200 MHz frequency range in MHz steps. At each excitation frequency the agnitude and phase of the detected signal was easured. The bandwidth of each easureent was reduced to Hz using the lock-in aplifier. Figure 3 shows the agnitude of acoustic signals on a 6 thick aluiniu block as a function of frequency, with a source to receiver distance of 256µ. The cobination of the Michelson interferoeter with lock-in detection allows for excellent sensitivity. The agnitude of the displaceent easured over this frequency range is in the fetoeter range. There is a fall-off in the signals at higher frequency due priarily to a reduced odulation depth in the excitation source. The odulation depth falls fro about 85% at MHz to 65% at 200MHz. It is unclear what is

5 responsible for the reduced aplitude at low frequencies, but we hypothesize that this ay be due to interference effects due to the waves reflecting off of saple boundaries. The data is processed in order to synthesize the tie doain response of the syste to a pulsed excitation source. The source function is given by:.00 Relative Aplitude Tie (ns) Figure 4. Synthesized input function used to obtain transient response of syste. i( t) = Pn ( nf0 )cos( 2. 0π n fot), () n= where f 0 is the frequency step, P n is the peak power of the excitation source at frequency nf 0, and is the total nuber of easureents taken. Note that Eqn. is siply a atheatical construction and does not represent the actual incident laser intensity, but rather a su of the excitation sources over all of the incident odulation frequencies. Once the response of the syste is known over a given frequency range, the output of the syste to any arbitrary input signal can be deterined. In the case that P n is constant over the entire frequency range and the frequency step is infinitely sall, Eqn. represents an sinc function centered at t = 0. Figure 4 shows a synthesized source function over the MHz frequency range with a frequency step size of MHz. The peak power at each frequency is taken to be constant and the output is noralized. The tie doain output of the syste s(t) is found by suing the signals at each of the easured frequencies: s( t) = M n ( nf0 )cos[ 2.0π n fot + θ n ( nf 0 )], (2) n= where M n (nf 0 ) is the agnitude of the signal easured at frequency nf 0 and θ(nf 0 ) is the corresponding phase. The DC coponent of the signal is not easured, and thus is not reconstructed in either the source or output signals. The data can also be inverted using an inverse FFT algorith rather than the Fourier series representation given above. Figure 5 shows the results of reconstructing tie doain signals using the frequency doain data acquired over 50, 00, and 200 MHz. The data is taken over MHz steps and thus

6 the nuber of wavefors sued increase for the wavefors reconstructed over the larger bandwidths. As expected, the signals becoe ore localized in tie as the bandwidth increases. In the signal reconstructed over 200 Mhz, both the surface skiing longitudinal wave (SSL) and surface acoustic wave (SAW) are observed. The advantage of using tie doain reconstruction is evident; individual acoustic arrivals can be identified and tie gated. This is particularly useful when ultiple acoustic odes are generated in the syste or when looking for a reflected or scattered acoustic signal MHZ Aplitude (f) SSL SAW 00 MHz 50 MHZ Tie (ns) Figure 5. Tie doain signals on an aluiniu plate reconstructed over bandwidths of 50, 00, and 200 MHz. Discussion: The use of CW generation, cobined with lock-in detection, allows for the detection of displaceents in the fetoeter range, generated with only oderate surface heating. We estiate that the axiu heating of the aluiniu surface used in these experients was less than 20 degrees. The cobination of this approach with tie doain reconstruction provides tie doain resolution and the ability to tie gate for signal processing an analysis. It is noted that the iproveent in signal-to-noise ratio does coe at the expense of easureent tie. For tie sensitive applications, a faster integration tie ay be required. However, it ay be possible to copensate for this in the inspection of soe aterials through the use of a higher power generation source, while still working well away fro the aterial ablation threshold. We also expect that this technique can be used for the easureent of fil thickness or echanical properties using surface acoustic waves. The frequency of the detection syste is currently liited to 200 MHz due to the RF lock-in response. We hope to extend this to the GHz range by either ixing the signal down to a lower frequency or detecting the signal with a vector network analyser. Conclusions: A laser based acoustic icroscopy syste has been developed which uses an aplified electroabsorption odulated laser source for generation. The syste provides a substantial iproveent in sensitivity over systes using pulsed lasers for generation through a reduction in signal bandwith using a lock-in detection schee. In addition, we deonstrate that it is possible to reconstruct the tie doain response of the saple, giving signals siilar to those observed using pulsed laser sources. This allows for individual acoustic arrivals to be easily identified and for tie gating. Experiental results deonstrate that the displaceent sensitivity of our detection syste is in the fetoeter range. Tie doain reconstructions of acoustic signals generated in an aluiniu block show the presence of both the surface skiing

7 longitudinal wave and surface acoustic wave. Future work will focus on increasing the axiu detection frequency and on using the syste for thin fil inspection. References:. C.B. Scruby and L.E. Drain, Laser Ultrasonics, Techniques and Applications, Ada Hilger, N.Y. (990). 2. S.J. Davies, C. Edwards, G.S. Taylor, and S.B. Paler, Laser-generated ultrasound: its properties, echaniss, and ultifarious applications, J. Phys. D (993). 3. J.P. Monchalin, C. Neron, J.F. Bussiere, P. Bouchard, C. Padioleau,R. Heon, M. Choquet, J.D. Aussel, G. Durou, and G. Nilson, Laser-ultrasonics: Fro the laboratory to the shop floor, Adv. Perfor. Mater. 5 (-2) 7-23 (998). 4. S.G. Pierce, B. Culshaw, and Q. Shan, Laser generation of ultrasound using a odulated continuous wave laser diode, Appl. Phys. Lett. 72, 030 (998)