Seismic characteristics of cavity decoupled explosions in limestone: An analysis of Soviet high explosive

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. B12, PAGES 27,393-27,405, DECEMBER 10, 1997 Seismic characteristics of cavity decoupled explosions in limestone: An analysis of Soviet high explosive test data J. R. Murphy, I. O. Kitov, N. Rimer, 3 V. V. Adushkin, and B. W. Barker Abstract. During the summer of 1960, Soviet scientists conducted a series of high-explosive, cavity d½coupling tests in a mine in Kirghizia. These d½couplcd tests were carried out in a variety of mined cavities in limestone, including spherical cavities with diameters ranging from 3.5 to 10 rn as well as nonspherical cavities with volumes of about 25 m 3. The experiments of this test series consisted of 10 tamped and 12 d½coupled explosions having yields of 0.1, 1.0, and 6.0 t, and seismic data were.recorded at locations in the mine over a distance range extending from about 10 to 250 m from the sources. These data provide valuable new insight into the dependence of d½coupling effectiveness on variables such as cavity volume, cavity shape, and charge ½mplac½mcnt geometry. In particular, analyses indicate that chemical explosions at a depth of 290 m in limestone are essentially fully decoupled in spherical cavities with scaled cavity radii larger than about 27 m/kt /3 and thathe low-frequency decoupling effectiveness under such conditions is approximately independent of cavity shape for roughly cylindrical cavities with length-to-width ratios of as much as These restilts suggesthat the possibility of cavity decoupling in nonspherical cavities in hard rock media will have to be carefully evaluated in establishing the seismic verification regime for the Comprehensive Test Ban Treaty. 1. Introduction It has long been recognized that the most effective means for evading the seismic detection of a clandestine underground nuclear test is to detonate the explosion in a cavity that is large enough to substantially decouple the radiated seismic signal. In fact, the cavity decoupling evasion scenario continues to representhe greatest challenge to reliable seismic verification of compliance with the Comprehensive Test Ban Treaty (CTBT). However, despite the fact that the feasibility of this evasion concept was experimentally established nearly 30 years ago by the U.S. STERLING nuclear cavity decoupling test, a number of issues of importance with respect to seismic monitoring still remain unresolved. This is the case because the available experimental database is too sparse and uncertain to provide a finn basis for resolving these questions and because the theoretical models being used to simulate the seismic source characteristics of such tests are strongly dependent on the poorly constrained low-pressure equations of state used to representhe response of real Earth materials to explosive loading [Murphy et al., 1996]. As a result, significant uncertainty is associated with extrapolations to cavity decoupling test conditions that are outside the range of previous experience. From a practical perspective the principal unresolved issues relate to the assessment of the effects of cavity size and Federal Division, Maxwell Tecbmologies, Inc., Reston, Virginia. 2Institute for Dynamics of the Geospheres, Russian Academy of Sciences, Moscow. 5Federal Division, Maxwell Technologies, Inc., San Diego, Califor- nia. Copyright 1997 by the American Geophysical Union. Paper number 97JB /97/97JB $09.00 shape on decoupling effectiveness. That is, because the spherical cavity sizes required for theoretically "full" decoupling in the yield range of potential testing interest are so large and difficult to construct, there is a significant incentive for a clandestine tester to evaluate quantitatively the penalties associated with using progressively smaller cavities and elongated cavities that are nonspherical in shape. Over the past several years we have carried out a wide range of nonlinear, finite dif- ference simulations of cavity decoupling in which the effects of cavity size and shape have been analyzed for both nuclear and chemical explosions in various source media [Murphy et al., 1988; Stevens et al., 1991a,b; Rimer et al., 1994; Murphy et al., 1996]. While the results of these theoretical simulations have been very informative, it has been difficult to assess critically their fidelity because of the limited U.S. experimental database on cavity decoupling. However, scientists from the Institute for Dynamics of the Geospheres (IDG) of the Russian Academy of Sciences have recently begun publishing new information on some Soviet cavity decoupling experiments [e.g., Adushkin et al., 1992] that provided relevant data. For this reason, scientists from Maxwell Technologies, h c., and IDG initiated a series of joint research investigations in which attempts have been made to integrate these newly available data and theoretical results in order to develop an improved, quantitative capability for evaluating the plausibility of various cavity decoupling evasion scenarios. h this paper we describe the results of our comprehensive analysis of seismic data recorded fi'om a Soviet series of high-explosive (HE) decoupling tests conducted in cavities of different sizes and shapes in limestone in a Kirghizia uranium mine in , Description of the Experiments During the summer of 1960, Soviet scientists carried out a series of HE cavity decoupling tests in limestone at a mine in the Tywya Motretains of Kirghizia (40.4øN, 72.6øE). This test

2 27,394 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE W - E DECO UPLiNG P TEST AREA MINE i loo I M MINE Figure 1. Vertical sections (top) along and (bottom) perpendicular to the mine access tinreel used for the Kirghizia cavity decoupling experiments. series was comparable in many ways to the COWBOY HE decoupling test series that was conducted in salt in the United States at about that same time [Herbst et al., 1961], although it was somewhat more comprehensive in that it included a number of charge configurations that were not investigated in the COWBOY tests. h particular, the Kirghizia series included tests designed to evaluate the effects of cavity shape and charge emplacement geometry on decoupling effectiveness, in addition to conventional spherical cavity tests similar to those employed in COWBOY. The tests were conducted in chambers that were excavated off the main mine access ttmnel at a depth of about 290 m be- low the surface. Figure 1 shows vertical sections along the access tinreel and perpendicular to the tunnel at the test location. It can be seen fi'om Figure 1 that the mine penetrates a mesa rather than a motretain and that the surface topography above the tests is relatively smooth over distances of the order of several hundreds of meters from ground zero. The relative locations and configurations of the various explosion chambers that were excavated for this test series are shown in Figure 2 where it cm be seen that the maximtun separation between any of the tests was less than 150 m. The five excavated decoupling test chmnbers shown here include three spherical cavities with dimneters in the 3.6 to 9.8 m range (i.e., radii of Chamber 6 Chamber 17 Chamber 7 Chamber 10 Chamber 13 Figure 2. Relative locations and configurations of the various explosion chambers constructed in the Kirghizia limestone mine for the Russian high-explosive (HE) decoupling test series.

3 MURPHY ET AL.: CAVITY DECOUPL1NG IN LIMESTONE 27,395 Chamber 7 rc = 1.81m 0.1 ton (39.0) 0.1 ton (39.0) Chamber 10 Chamber 13 Chamber 17 Chamber 6 rc =2.88m rc =4.92m v=25m 3 v=23m ton D 0.1 ton ff 0.1 ton!_.., ß 0.1 tond l l (62.0) 1.0 ton (49.2) 1.0 ton ton (49.2) (28.8) ton (27.1) 0.1 ton! ' (39'1) u Figure 3. Graphical summary of the Kirghizia HE decoupling tests conducted in each of the excavated explosion chambers. Asterisks denote the emplacement location of the charge within the chamber for each test. For the nonspherical cases (chambers 17 and 6) both (left) horizontal and (right) vertical sections are displayed. The numerical values in parentheses, below the yield values, are the scaled cavity radii in m/kw 3, with equivalent volume spherical cavity values listed for the nonspherical cases. 1.81, 2.88, and 4.92 m)as well as two nonspherical chambers of roughly cylindrical shape encompassing volumes of about cavity walls, so some of the cavities were reused. Multiple tests in a given chamber were always conducted in order of 25 m 3, approxi nately equal to that of the 1.81 in radiu spheri- increasing yield. It will be shown later that the seismic cal cavity. The shapes of these two nonspherical cavities were quite different. Chamber 6 was roughly cylindrical with a length L of about 6 m and height H and width W of about 2 rn, giving an effective aspect ratio of about 3. Chamber 17, on the other hand, was considerably more elongated (L = 12 m) and decoupling estimates for these multiple test cases are very consistent, which strongly supports the hypothesis that the dynamic response characteristics of the cavity walls were not significantly altered by the prior, lower-yield tests. Radial component seismic data were recorded froin these asymmetrical in cros section (W = 2 rn, H = 1 m), correspond- Kirghizia tests at shot depth in the mine over a distance range ing to an equivalent aspect ratio in the range of 6 to 12. Thus extending from about 10 to 250 n (Kirov et al., 1995). It foldata recorded from the tests in these chambers can provide a lows that the azimuthal distribution of the recorded data is good measure of the dependence of decoupling effectiveness somewhat limited by the mine gemnetry. However, as will be on cavity shape over a fairly wide range of cavity aspect ratios. demonstrated in the following analysis, the consistency of the These cavities were excavated in hard, homogeneous limestone, characterized by compressional wave velocities in the 5.5 to 6.0 krn/sec range. The inference of homogeneity is based Soviet reports of visual observations of the media exposed in recorded data between these various tests provides strong evidence that the seismic sources are essentially isotropic, at least for the spherical cavity tests. Most of these data were recorded on broadband velocity sensors (VIB-A) that have an essentially flat response over the 10 to 500 Hz band analyzed the tunnel walls and in the excavated cavities as well as the consistency of the reported first arrival times as ftmctions of range for the various tests. The test series was composed of 10 tamped and 12 decou- pled explosions with yields of 0.1, 1.0, and 6.0 t. The explosives consisted of ammonimnitrate, except for the two 6.0 t tests that utilized a mix of TNT and ammonium nitrate. For the cavity tests the explosives were suspended in the chambers and included cases in which the explosives were positioned in the center of the cavity as well as cases in which they were positioned off center, near the cavity walls. The configurations of the various cavity tests are graphically summarized in Figure 3 for each of the five test chambers. It can be seen that explosions of the same yields were detonated in cavities of different size and also that explosions of different yields were detonated in two of the chaxnbers (i.e., 10 and 13), thus providing redtmdant data that can be used to assess the of variations in scaled cavity size on decoupling effectiveness. It was originally pla med to use each chamber only once, but postshot visual inspections revealed no detectable damage to the in this study. These sensors were installed in drill holes and niches excavated in the wall of the mine. At locations where both eraplacement techniques were employed, no significant differences in signal characteristics were observed for these relatively low-level seismic motions. Peak amplitudes of displacement and velocity were determined for more than 250 of these recording locations by the Soviet scientists involved in the original experiment, and these peak data are analyzed and cronpared in section 3. Unforttmately, some of the original data were lost in the insuing 35 years and waveform data were recovered for only about 30 of these recordings for use in this study. These recovered waveform data were digitized at IDG and form the basis for the subsequent frequency dependent decoupling analysis. Before moving on to the analysis of the data, it is appropriate to consider how the yield/cavity volume ratios for these Kirghizia tests correspond to the common reference values of this ratio. Now, according to the simplified Latter criterion [Latter et al., 1961 ], the volume of the cavity required to

4 27,396 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE decouple an underground explosion is directly proportional to the yield of the explosion and inversely proportional to the overburden pressure at shot depth. It follows that the cavity radius required to decouple a nuclear explosion of yield W at a depth of 290 m in Kirghizia limestone to the same degree as that achieved for the 0.38 kt STERLING explosion in a 17 m radius cavity at a depth of 828 m in salt is given approximately as 32W m m, for W in kilotons. Thus the equivalent STERLING cavity radii for 0.1, 1.0, and 6.0 t nuclear explosions at 290 m in limestone are about 1.5, 3.2, and 5.8 m, respectively. It follows that for HE/nuclear equivalence ratios in the range of 1 to 2, the 0.1 t Kirghizia HE tests in the 1.81, 2.88, and 4.92 m radius cavities and the 1.0 t test in the 4.92 m radius cavity should have been decoupled at least as effectively as STERLING, while the 1.0 t test in the 2.88 m radius cavity and the 6.0 t test in the 4.92 m radius cavity are somewhat overdriven with respect to STERLING, at least according to the Latter criterion. These highly si nplified calculations provide a rough basis for evaluating the Kirghizia li nestone data in tenns of previous experience in salt. 3. Analysis of the Peak Amplitude Data The peak displacement data observed f om the 0.1, 1.0, and 6.0 t tamped explosions are plotted as a function of source/receiver distance in Figure 4. These tmnped tests were all conducted within 100 m of the corresponding cavity tests in alcoves excavated off the mine access tinreel. Multiple tests were conducted at the 0.1 and 1.0 t yields, and it can be seen that the observed data from these tests are reasonably consistent and that they provide well-constrained average amplitude levels as a ftmction of distance over the range extending from about 10 to 200 m A least squares statistical analysis of the extensive 0.1 t tamped data set gives m average distance decay D p, gm The peak displacement amplitude data observed from the 1.0 rate that is approximately proportional to R ' ] over this range. t tamped test of Figure 4 are compared in Figure 5 with the corresponding peak displacement data as a fimction of range observed fzom the 1.0 t &coupled test in the 2.88 n radius 104. ton spherical cavity. It can be seen from Figure 5 that the cavity test was indeed &coupled in this case, showing an average 1.0 ton peak displacmnent decoupling factor of about 10 with respect to the tamped shot of the same yield. Note that this value is significantly lower than the nominal low-frequency decou pling factor of about 70 which is generally associated with the fully decoupled tests of the U.S. STERLING and COWBOY 0.1 series. However, it is importanto note that the peak dis- Dp, gm placements compared in Figure 5 correspond to very different o 102- frequency components and therefore cannot be directly interpreted in terms of the low-frequency decoupling level that is typically used to quantify decoupling efficiency. This fact is clearly illustrated in Figure 6 which shows a comparison of the radial particle velocity seismograms recorded at a range of about 100 m from these two explosions. Note that these data are consistent with the theoretically expected differences in dmninant frequency content between tmnped and cavity Dp~r -1.1 &coupled explosions of the stone yield. That is, because the characteristic seismic source radius is considerably larger for a 100 tamped explosion than for a cavity &coupled explosion of the 0 o 10 1'0 2 stone yield, the characteristicomer frequency of the tmnped r, rn seis nic source is expected to be lower, as observed in Figure Figure 4. Comparison of peak displacmnent data observed 6. It follows that spectral analyses of such complete waveform from the various tamped HE tests that provide the reference data are required in order to accurately esti nate the absolute base for the decoupling analysis. levels of the low-frequency decoupling factors for these tests o W = 1.0 Torl ß Tamped Dp = 2.3 x 104R '1.1 ( Decoupled, r c = 2.88 m Dp = 2.0 x r, m Figure 5. Comparison of peak displacement data as a function of range observed from 1.0 t tmnped and cavity decoupled H E tests at Kirghizia. Consequently, the distance attenuation was constrained to be R ']' in the statistical analyses of the observedata frown the 1.0 and 6.0 t tmnped explosions, leading to the three parallel straight line fits to the data shown in Figure 4. It can be seen that all these data are quite consistent with the single, nominal distance attenuation rate.

5 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE 27,397 Tamped 1.0 ton cm/sec Decoupled 1.0 ton r c = 2.88m cm/sec Decoupled 1.0 ton r c = 4.92m... cm/si c I.01 sec I Figure 6. Comparison of radial particle velocity seismograms recorded at a range of about 100 m from 1.0 t tamped and cavity decoupled Kirghizia explosions. However, despite the fact that peak amplitude data are not well suited for establishing absolute levels of decoupling effectiveness, they do provide a reasonable basis of comparifroln 1.0 t decoupled explosions in cavities with radii of 2.88 and 4.92 txt It follows that peak amplitude readings obtained from such recordings can be directly compared to estimate difson that can be used to assess the relative effects of variables ferences in decoupling effectiveness at a common dominant such as yield, cavity size and shape, mad charge emplacement frequency. For example, Figure 7 shows a comparison of the geometry. That is, while the dominant frequencies of the peak motions corresponding to tamped and cavity decoupled explosions of the same yield are observed to be quite different, those associated with decoupled explosions of the same yield in different cavities are observed to be very similar. This fact is also illustrated in Figure 6, by way of a comparison of the particle velocity seismograms recorded at a range of about 100 m peak displacement data as a ftmction of range observed from 0.1 t decoupled tests in the spherical cavities with radii of 1.81, 2.88, and 4.92 m. Now, according to elementary theory, once the cavity is large enough to fully decouple an explosion of a given yield, the low-fiequency level of the seismic source ftmction is expected to be independent of cavity radius for any larger cavities [Murphy, 1980]. That is, in the low-frequency ¾o D o 102 D p,!m 10 0 '' [ Dp,m O r c = 1.81 rn ß r c = 2.88 rn [] r c = 4.92 rn 10 ø r, rn Figure 7. Comparison of peak displacement data as a function of range observed from 1.0 t decoupled tests in spherical cavities with radii of 1.81, 2.88, and 4.92 m. 0 r c = 2.88 rn ß rc = 4.92 m 10 ø,... i... I ø r, rn Figure 8. Comparison of peak displacement data as a function of range observed from 1.0 t decoupled tests in spherical cavities with radii of 2.88 and 4.92 m.

6 27,398 MURPHY ET AL.: CAVITY DECOUPLING LIMESTONE limit, the seismic source level corresponding to the linear, elastic response of the cavity wall to a late time, steady state pressure P in the cavity is proportional to Pro 3, and since P is inversely proportional to cavity volume, this source level is independent of cavity radius once the linear, elastic limit has been reached. It can be seen fi-om Figure 7 that the observed 102 _ peak displacement levels for these three 0.1 t tests appear to be independent of cavity radius over this range, which suggests that essentially full decoupling was achieved in all three of these tests. This is not surprising in that, as noted previously, Dp,m all three of these cavities are larger than that of STERLING, scaled to this yield and overburden pressure. It does, however, serve to confirm the fact that such low-frequency peak dis- placement data are useful for comparing the relative seismic efficiency of different decoupling tests. A similar comparison for the 1.0 t decoupled tests in the O (W=l.0ton) centers of spherical cavities with radii of 2.88 and 4.92 m is ß (W=0.1 ton)x 10 presented in Figure 8. In this case the 2.88 m radius cavity is [] (W = 6.0ton)/6 predicted to be overdriven somewhat with respect to ,,,,,,...,... STERLING, while the 4.92 m radius cavity test is expected to 10 ø be fully decoupled. However, once again, the observed peak r, rn displacement levels from the two tests appear to be quite com- Figure 10. Comparison of yield-scaled peak displacement data parable, suggesting that both tests were essentially fully as a function of range for 0.1, 1.0, and 6.0 t decoupled tests in decoupled. the spherical cavity with radius 4.92 in. Another way of comparing these same data is to look at scaled peak displacements for explosions of different yields in the same cavity. That is, by the same simple, elastic theory referenced above, the low-frequency level of the seismic source range observed from the 0.1 and 1.0 t explosions in that cavity function is expected to be directly proportional to yield for are coinpared. It can be seen that these yield-scaled peak discavities large enough to fully alecouple the explosions. placement values are very consistent, providing further evi- Therefore, to the extent that the observed peak displace]nent dence that the 1.0 t explosion in this cavity was essentially data are proportional to the low-frequency levels of the corre- fully alecoupled. Figure 10 shows a silnilar comparison of sponding seismic source ftmctions, they should scale as the yield-scaled peak displacements as a ftmction of range for the first power of the yield for fully alecoupled explosions in a 0.1, 1.0, and 6.0 t alecoupled tests in the center of the 4.92 in given cavity. The results of applying this model to the tests in radius spherical cavity. Here, again, the yield-scaled peak the 2.88 m radius spherical cavity are presented in Figure 9 displacement levels are found to be in excellent agreement, where the yield-scaled peak displacements as a function of indicating that all three of these tests were fully decoupled. We conclude that the available peak amplitude evidence indi- 10 3, 10 2' Dp,m O (W=l.0ton) ß (W=0.1ton)X r, rn Figure 9. Comparison of yield-scaled peak displacement data as a function of range for 0.1 and 1.0 t decoupled tests in the spherical cavity with radius 2.88 m cates that all the tests detonated in the center of these three spherical cavities were essentially fully decoupled. These results are summarized in Figure 11 which shows the yieldscaled peak displacement levels for the different cavity tests, plotted as a function of scaled cavity radius (rc/wi/3). For the purposes of this comparison, the relative peak displacement levels were esti]nated by computing least squares mnplitude/distance relations for each test, asstuning the nominal attenuation rate of r -1'1, corresponding to the slope of the straight lines on Figures It can be seen fi-o]n Figure 11 that these yield-scaled peak displacement levels show no obvious trend as a fimction of scaled cavity radius over the rm ge extending frotn 27 to 106 m/kt /3. The average scaled (i.e., W- 1 t, r = 1 m) peak displace nent level for these cavity tests is 2800 gm, with a total range of only about +25% around this mean value. We conclude that HE tests in spherical cavities with radii larger than 27 n/kt /3 trader this overburden pressure in Kirghizia limestone are essentially fully decoupled. The effects of charge emplacement geometry are addressed in Figure 12 which shows a comparison of the peak displacement data as a ftmction of range observed fi'om the two 1.0 t tests conducted at different locations in the 4.92 m radius spherical cavity. In this case, one test was conducted with the charge positioned in the center of the cavity, while the other was

7 MURPHY ET AL.: CAVITY DECOUPLING 1N LIMESTONE 27,399 E Figure 11. levels as a function of scaled cavity radius for Kirghizia decoupled HE tests in the center of spherical cavities at a depth of 290 m in limestone. elongated cavities are very comparable to those observed from the test in the spherical cavity of the same volrune over the entire distance range of observation. These results provide strong evidence that the low-frequency decoupling effectiveness is approximately independent of cavity shape for elongated cavities with aspect ratios of as much as Figure 13, right, shows a similar comparison for the test detonated 1 m from the end of chamber 17. ha this case the closest observa- tions at ranges less than about 10 m show evidence of some e hanced coupling relative to the corresponding spherical cavity observations, but, once again, these differences appear to be less than a factor of 2. Moreover, although the data are sparse, even these relatively small differences seem to disappear at observation distances greater than the long dimension of the cavity. This is a somewhat surprising result, given the corresponding spherical cavity observations of Figure 12, and suggests that this charge emplacement geometry issue deserves additional study. Of course, all of these nonspherical i cavity results have to be qualified by the uncertainties intro- 102 duced by the limited azimuthal coverage provided by the data rc /3 recorded in the mine. However, given the consistency of the wl/3' m/ktl data from the various tests shown in Figure 13, we provisionally conclude that over the range of test conditions explored at Comparison of yield-scaled peak displacement Kirghizia, the low-frequency decoupling effectiveness depends only on cavity volume and is roughly independent of the shape of the cavity, in agreement with our previous theoretical simulation results [Stevens et al., 1991b; Rimer et al., 1994]. conducted with the charge centered 1 m from the cavity wall. It 4. Analysis of the Waveform Data can be seen from Figure 12 that the observed peak displacement level for the test near the cavity wall appears to be some- It was noted above that the waveform data currently availwhat larger, on average, than that observed from the corre- able from the Kirghizia decoupling test series are much less sponding test in the center of the cavity. This suggests that complete than the peak motion data. In particular, of the more the proximity of the charge to the cavity wall in the former test than 250 recordings from Which peak motion values were deresulted in an increase in the degree of nonlinear response in the surrounding medium and hence increased seismic coupling efficiency in that case. That is, the decoupling efficiency has 103. been somewhat reduced as a result of this charge eraplacement geometry. However, the magnitude of this effect appears to be no more than a factor of 2 in this case, at least in the frequency range represented by this peak displacement data. As was noted previously, in addition to the spherical cavity decoupling tests described above, the Kirghizia series included several tests designed to assess the influence of cav- 102' ity shape on decoupling effectiveness. This is a very i nportant practical issue in that, from an engineering perspective, it is D much easier to construct elongated, tinreel-like cavities than it p,!.lm is to construct traderground spherical cavities of the same volume. A number of theoretical studies of this problem have been conducted in recent years (Stevens et al., 1991b; Rimer et al., 1994), and the results have indicated that the lowfrequency decoupling effectiveness is largely independent of cavity shape, even for elongated cavities with aspect ratios of 10 to 1 or more. However, until now, no experimental data have been available to test these theoretical simulation re- sults. Figure 13 shows comparisons of the peak displacement versus range data observed from the Kirghizia 0.1 t decoupled tests in spherical and cylindrical cavities of about the same volume (i.e., 25 m3). Figure 13, left and middle, show comparisons for the explosions detonated in the center of test chambers 6 and 17 (see Figure 3), respectively. It can be seen that the peak displacement levels observed from these two tests in ,,,,..., 10 ø 10 2 r, ITI Figure 12. Comparison of peak displacement data observed from 1.0 t decoupled tests at different locations in the 4.92 m radius spherical cavity.

8 27,400 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE ' r c = 1.81m 0 r c= 1.81m, I'- lv = 23rn3, UW = 3, I" V = 25rn3, UW = 6-12 Y [' 'l v = 25rn 3, UW = 6-12 I... io I rc= 1.81m 10 o 102 r, rn r, rn r, rn Figure 13. Comparison of peak displacement data observed from Kirghizia 0.1 t decoupled tests in spherical and elongated cavities of comparable volume. For the nonspherical cavities, asterisks denote the charge loca- tion. termined, waveform data were recovered from only about 30, all of which have now been digitized at IDG and previewed for possible spectral analyses. Unfortunately, more than half of these waveforms were found to be complete through only the first half cycle of motion, and, although adequate for determination of peak amplitudes, these proved to be of limited value for quantitative evaluation of the frequency dependent decoupling on these tests. Moreover, of the remaining complete waveforms a number were recorded from 0.1 t tamped tests for which there are no corresponding recordings from cavity decoupled tests of the same yield. Thus the only waveform data that proved to be useful for quantitative spectral analyses were the few recorded from 1.0 t tamped and decoupled tests in the 2.88 and 4.92 m radius spherical cavities. The radial component particle velocity seismograms recorded in the distance r,-mge of 77 to 193 m from two differentamped 1.0 t Kirghizia tests are plotted in Figure 14, while the corresponding seismograms recorded from two 1.0 t cavity decoupled tests are reproduced in Figure 15. All these data were digitized at rates exceeding 10,000 samples per second. Figure 15, top trace, was recorded from the 1.0 t explosion in the center of the 2.88 m radius cavity, while Figure 15, middle and bottom traces, r= 77m 10 cm/sec r = 117rn 4 cm/sec r= 141m 5 cm/sec r = 193m.01 sec I 4 cm/sec Figure 14. Radial component particle velocity seismograms recorded from 1.0 t tamped Kirghizia explosions.

9 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE 27,401 1 ton, r c = 2.88m r = 97m cm/sec 1 ton, r c = 4.92m r = 85m cm/sec 1 ton, rc = 4.92m r = 110m =!, cm/sec.01 sec Figure 15. Radial component particle velocity seismograms recorded fi-om 1.0 t decoupled Kirghizia explosions in the (top) center of a 2.88 m radius spherical cavity and 1.0 m froin the wall of a 4.92 m radius spherical cavity for r of (middle) 85 m and (bottom) 110 in. were recorded froin the 1.0 t explosion detonated 1 in fi-om the is critically dependent on the assumption that this one recordwall of the 4.92 m radius cavity. It can be seen that these three ing is representative of that test. ha the absence of additional decoupled test waveforms were all recorded at a range of about data, about all that can be said in this regard is that the peak 100 m, a distance that is encompassed by the ranges of the four displacement value corresponding to this recording falls right tamped recordings of Figure 14. Spectral analyses of these complete waveform data indicate that, in all cases, the signal- 102 to-noise ratio is greater than a factor of 3 over the entire analyzed frequency band extending from about 10 to 500 Hz. Although detailed analyses of the spectra corresponding to the tamped waveforms of Figure 14 revealed that the amplitude attenuation with distance in the Kirghizia limestone is frequency dependent, as would be expected, the decoupling analysis results were found to be insensitive to the details of this frequency dependence because the available recorded data sample such a liinited distance range. Consequently, we proceeded by normalizing all the data to a reference distance of 110 in using the nominal, frequency independent peak veloc- 0 ity attenuation law r '1'75. Following this distance normalization, tamped to decoupled spectral ratios were computed using 100 each of the four tamped recordings, and the results were then averaged to estimate the decoupling factors as a function of frequency. The result for the 1.0 t explosion in the center of the 2.88 m radius cavity is shown in Figure 16 where it can be seen that the maximum low-frequency decoupling at the comer frequency of the tamped shot (i.e., aromad 75 Hz) is about a 10-1 factor of 25 for this HE test. The dotted lines in Figure 16 de- note the 95% confidence interval around the mean spectral f, Hz 102 ratio, and it can be seen fi'om these bounds that the frequency dependent decoupling estimates obtained using the four dif- Figure 16. Frequency dependent decoupling factors corresponding to 1.0 t Kirghizia explosions in the center of a 2.88 ferent tamped recordings are quite consistent, particularly at m radius spherical cavity (solid line) and near the wall of a low frequencies. This measure of uncertainty is limited, of 4.92 m radius spherical cavity (dashed line). The dotted lines course, by the fact that only one recording is available from the denote the 95% confidence interval about the mean spectral decoupled test. Thus the accuracy of the decoupling estimate ratio for the 2.88 m radius decoupled explosion.

10 27,402 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE on the least squares fit to the data recorded from this test (see estimate the seismic source characteristics of selected tests. At Figure 8), which suggests that it is indeed representative of the decoupled seismic source. The observed maximum low-frequency decoupling factor of 25 for this presumably fully decoupled Kirghizia test is conthis time, no material properties data are available for the rock strength at the Kirghizia site, and this is unfortunate because rock strength plays a key role in defining seismic coupling efficiency. In particular, for tamped events, higher strength siderably lower than the nominal full decoupling factor of 70 results in lower displacements and lower seismic coupling that is usually quoted on the basis of the STERLING experi- efficiency. For relatively decoupled cavity events the strength, ence. Although a number of recentheoretical simulation stud- particularly at low mean stresses corresponding to the preshot ies have indicated that for a given cavity size, HE explosions in situ rock stress, determines the scaled cavity radius for full are expected to decouple less effectively than nuclear explosions of the same yield in salt [Glenn and Goldstein, 1994' Murphy et al., 1996], the observe decoupling factors for the fully decoupled COWBOY HE tests appear to have been about 70 [Murphy, 1980], close to that reported for STERLING. However, very little decoupling data have been reported from tests in media other than salt. Reinke et al. decoupling. Strength values at higher pressures may be expected to be relatively consistent from site to site. However, the strength at low pressure can vary dramatically in the same rock type fxom site to site as a result of microfracturing and macrofracturing patterns in the rock. Since both the 2.88 and 4.92 m cavities at Kirghizia appear to be fully decoupled for a 1.0 t explosive yield, these data provide a lower bound to the [1995] described recently some preliminary results obtained strength of the limestone at lower pressures. h the absence of from the analysis of data recorded flora cavity decoupled HE tests conducted in limestone in a mine near Magdalena, New Mexico. h these tests, explosions with a wide range of yields (i.e., 0.1 to 4.0 t) were detonated in rectangular chambers of fixed dimensions (i.e., 2 m x 4 m x 8.5 m). These tests were somewhat limited, however, in that these chambers were not sealed and the explosive pressures were vented to the main any other observational constraints it has been assumed that the physical properties of the Kirghizia limestone are similar to those reported in the literature for other hard limestone formations [Chitty and Blouin, 1993]. The results of interest are, in general, less sensitive to parameters of the equation of state (EOS) of the explosive detonation products. The chemical explosive used for the Kirghizia mine adit via a narrow access passageway to the chamber. Thus 1.0 t events was described as ammonium nitrate with a density the applicability of these decoupling data is subjecto some of roughly 1.6 gm/cm 3. Since this explosive has not been well tincertainty, particularly at the low frequencies of primary in- characterized, most of the simulations were made for the wellterest. h any case, Reinke et al. [1995] report 1ow-fcequency characterized TNT explosive that has a density of 1.63 gngcm 3. decoupling factors for the Magdalena tests that range from The JWL (Jones-Wilkins-Lee) EOS for high explosives [Lee et about 10 to 70 for those cavity tests with scaled yield/volume al., 1973] was used to model the explosive detonation and to ratios, comparable to that of the 1.0 t Kirghizia test in the 2.88 m radius spherical cavity. Thus these two limestone HE compute the cavity pressure variation with time. Material properties for TNT for this EOS are given by Dobratz [ 1981 ]. decoupling estimates are consistent within the rather large Several calculations were also made using a lnodified TNT data scatter. A spectral m alysisimilar to that described above for the 2.88 m radius cavity test was also conducted using the two waveforms from Figure 15 that were recorded from the 1.0 t EOS, intended to better simulate the mnmoxfium nitrate (AN) explosive. This AN simulant had the same initial density, a much lower Chapman-Jouguet pressure (1.21 x 104 MPa) compared with TNT (2.1x104 MPa) and a lower detonation velocdecoupled test detonated near the wall of the 4.92 m radius ity, 5270 versus 6930 m/s, for TNT. In general, calculations cavity. The resulting average, frequency dependent decou- with this AN model gave decoupling factors (tamped/cavity) pling factor for this test is shown as a dashed line in Figure that were very similar to those calculated using TNT. How- 16, where it is compared with the corresponding average ever, the AN decoupling factors did not show the complicated decoupling factor for the 2.88 m radius cavity test. It can be seen that the maximum observed low-frequency decoupling for this test is about a factor of 10, which is 2.5 times lower than that observed for the 1.0 t decoupled test in the center of the 2.88 m radius cavity. That is, these spectral data are generally consistent with the peak displacement data of Figure 12, that were interpreted to indicate that proximity to the cavity wall structure at frequencies of Hz that were seen for the TNT results to be discussed later. For the tamped cavity case the computed low-frequency seismic coupling efficiency for the AN simulant was roughly 10% smaller than for the TNT case. For the decoupled cavity events, shock wave propagation in the air surrounding the chemical explosive was modeled had increased the low-frequency seismic coupling efficiency using an air EOS originally developed at the Air Force Weapby about this amount for that test. Note also flora Figure 16 ons Laboratory. This EOS has been used in a number of earlier that this difference in decoupling effectiveness is frequency decoupled cavity studies such as that by Stevens et al. dependent and that it reaches a max/mum of more than a factor [1991a]. For the simulations of decoupled nuclear events the of 4 at around 80 Hz, where the decoupling is maximum for the explosive energy was uniformly distributed in a volume of air 2.88 m radius cavity test. Thus the effects of charge emplace- large enough to produce an initial pressure of roughly 1.0 to ment geometry appear to be complex, m d additional theoretical simulation studies and experimental tests are required in order to define these effects in a quantitative fashion. 1.5x10 MPa using this equation of state. However, for the tamped nuclear case the initial cavity in the calculations was made large enough to vaporize roughly 70 t of rock per kiloton of explosive yield. Using an ideal gas equation of state for 5. Theoretical Simulation Analysis In tu attempt to derive a better quantitative understanding of the Kirghizia decoupling test results, some preliminary, nonlinear, finite difference simulations have been conducted to the cavity materials, this resulted in an initial cavity pressure of6.5x104 MPa. The calculated seismic source functions were forrod to be less sensitive to these simplifying assumptions than to the poorly known material properties of the limestone site.

11 MURPHY ET AL.' CAVITY DECOUPL1NG IN LIMESTONE 27,403 Least Squares Theoretical 10 3: 1ø2, X ß o o't : c 0o ø',% O' ,, 10'1...,..., ø ø r,m r,m Figure 17. Comparison of simulated and observed (left) peak displacements and (right) velocities for the 1.0 t Kirghizia decoupled test in the center of the 2.88 m radiuspherical cavity. Spherically symmetric, finite difference calculations were served r '1'75 attenuation rate is inconsistent with the theoreticarded out to simulate the seismic source functions corre- cal elastic solution and must be associated with some anelassponding to the 1.0 t Kirghizia tests for the tamped and 2.88 m radius cavity configurations. For this m d all subsequent cavity simulations the Lam6 elastic solution was used in the calculations to insure that the decoupled cavities were in equilibrium with the 7.7 MPa overburden pressure at the onset of the simulations. For the material model described above, the limestone surrounding the 2.88 m radius cavity did not fail in shear or tension, and, consequently, this case provides a good test of the low-pressurequation of state employed in these simulations. The peak displacements and velocities,as a timetlon of range resulting from this cavity simulation are plotted in Figure 17 where they are compared with the observedata from this test. It can be seen that the simulated peak displacements lie somewhat below the data in this case, although the tic decay mechanism that is not modeled in these calculations. However, even in the absence of this discrepancy between theoretical and observed attenuation with distance, it is clear from the close-in comparison at a range of 10 m that the theoretical simulation is significantly overestimating the observed peak velocity levels for this test. Similar comparisons for the 1.0 t tamped test are presented in Figure 18 where, once again, the calculations are seen to be in better agreement with the observed peak displacements than with the corresponding observed peak velocities. This tendency for the simulations to underestimate the peak displacements and overestimate the peak velocities suggests that the limestone material model being employed in these calculations has a shear strength that is too large. However, the reasonable predicted and observed attenuation rates are in good agree- agreement between the calculated and observed peak disment. The corresponding computed peak velocities, on the placements for these two tests suggests that this model should other hand, lie well above the mean of the observed data and provide an adequate description of the low-frequency decoushow a significantly lower rate of attenuation with distance. pling efficiency, which is of primary interest for seismic moni- In fact, if this test was indeed fully decoupled, then the ob- toring purposes. Least Squares Theoretical E 10 4 ß 103, Q 'o, 10o/ 10 ø ø r,m 102 Figure 18. Comparison of simulated and observed (left) peak displacements and (right) velocities for the 1.0 t Kirghizia tamped test. r,m 103

12 27,404 MURPHY ET AL.' CAVITY DECOUPLING IN LIMESTONE 10 2 CL O (,3... Theoretical HE... Theoretical NE ,,, 102 ß overestimates the comer frequency of the tamped test by a significant amount. This result is consistent with the peak ]notion comparison shown in Figures 17 and 18 and provides further evidence that the limestone ]nodel being used in the theoretical simulations is too strong. The theoretical simulation also fails to reproduce the observed decrease in decoupling efficiency with decreasing frequency below the comer frequency. This suggests that the seismic source of the tamped explosion shows overshoot of the low-frequency asymptotic value that is not matched by the theoretical simulation, which is also consistent with the hypothesis that our limestone model is too strong. It follows that additional work on the equation of state of limestone will be required in order to define a completely satisfactory simulation capability for applications to the higher-frequency characteristics of the seismic sources for such tests. The dotted line in Figure 19 shows the corresponding frequency dependent decoupling predicted for a 1.0 t nuclear source in this same cavity. It can be seen that for this source the maximran low-frequency decoupling factor is predicted to fall in the range of 50 to 60, which is significantly larger than the corresponding HE decoupling factor of 25 and close to the nominal factor of 70 quoted for the STERLING experiment. f, Hz This difference between predicted decoupling factors for HE Figure 19. Comparison of observed and theoretical frequency and nuclear sources of the same yield in a fixed cavity has been dependent decoupling factors for 1 t explosions in a 2.88 m noted before from theoretical simulations of decoupling in salt radius spherical cavity in Kirghizia limestone. [Glenn and Goldstein, 1994' Murphy et al., 1996]. Figure 20 shows a comparison of our predicted low-frequency decoupling factors for HE and nuclear sources in limestone and salt, The simulatedecoupling as a fimction of frequency for the plotted as a function of scaled cavity radius up to the scaled 1.0 t HE test in the center of the 2.88 m radiuspherical cavity radii where full decoupling is predicted. It can be seen that the is shown in Figure 19 where it is compared with both the cor- theoretically predicted differences between the HE and nuclear responding experimental estimate from Figure 16 and the theo- decoupling factors are qualitatively similar for these two rneretical prediction for a 1 t nuclear explosion in that same cav- dia, although the absolute levels of decoupling efficiency do ity. It can be seen thathe theoretical HE simulation predicts a appear to show some medium dependence. maximum low-frequency decoupling factor of about 25 in this case, in good agreement with the estimate obtained from the measuredata on that test. The agreement between the simu- 6. Summary and Conclusions lated and observed frequency dependence of the decoupling factor is less satisfactory in that it appears that the simulation The analysis of the seismic data recorded froin the Kirghizia HE decoupling test series has provided much new information Limestone Salt 10 2 ß - 10 o o NE --- HE NE --- HE lo 0 lo row-l/3 m/kt 1/31ø2 01, row-l/3, m/kt 1/3 02 Figure 20. Comparison of simulated low-frequency decoupling factors for HE and nuclear explosions in (left) limestone and (right) salt, plotted as functions of scaled cavity radius.

13 MURPHY ET AL.: CAVITY DECOUPLING IN LIMESTONE 27,405 regarding the dependence of decoupling efficiency on scaled cavity size and shape for explosions in a hard rock medium. In particular, the results of this analysis indicate that HE explosions at a depth of 290 m in limestone are essentially fully decoupled in spherical cavities with scaled cavity radii larger than about 27 m/kt 1/3 and that low-frequency decoupling effectiveness under such conditions is approximately independent of cavity shape for roughly cylindrical cavities with length-to-width ratios of as much as Spectral analyses of the limited, available waveform data indicate a maximum lowfrequency decoupling factor of about 25 for fully decoupled HE tests, and theoretical simulation results suggest that the corresponding nuclear decoupling factor would be expected to be in the range of 50 to 60. Thus these experimental data support our previous theoretical simulation results regarding the feasibility of effectively decoupling low-yield nuclear explosions in nonspherical cavities in hard rock and indicate that such evasion scenarios will have to be carefully evaluated in establishing the seismic verification regime for the CTBT. Acknowledgments. The authors thank J. L. Stevens and K. H. Lie tbr their assistance in carrying out the finite difference simulations. We would also like to acloxowledge A. N. Romashov, the chief of the expedition that carried out the tests, and D. D. Sultmxov of IDG for helpful discussions. This research was sponsored by the U.S. Air Force Phillips Laboratory under contract F C References Adushkin, V. V., D. D. Sultanov, I. O. Kitov, and O. P. Kuznetsov, Results of Experimental Study of Seismic Efficiency of an Explosion in an Underground Cavity, Rep. Rus. Acad. Sci., 324 (5), of Moscow, Chitty, D. E., and S. E. Blouin, Laboratory investigation of the strength and deformation properties of carbonate rocks and soils, Rep. DNA- TR-92-45, Defense Nuclear Agency, Alexandria, Va., Sept Dobratz, B. M., LLNL explosives handbook, properties of chemical explosives and explosive simulants, Rep. UCRL-52997, Lawrence Livem ore Natl. Lab., Livermore, Calif., March 1981 Glenn, L. A., and P. Goldstein, Seiõmic decoupling with chemical and nuclear explosions in salt, J. Geophys. Res. 99, 11,723, Herbst, R. F., G. C., Werth, and D. L. Springer, Use of large cavities to reduce seisnfic waves from underground explosions, J. Geophys. Res., 66, 959, Kitov, I. O., D. D. Sultanov, V. V. Adushkin, V. N. Kostuchenko, O. P. Kuznetsov, P. B. Kaazik, N. I. Nedoshivin, and H. D. Rubinshtein, Analysis of the seismic characteristics of U.S. and Russian coupled and cavity decoupled explosions in salt, Rep , Inst. for the Dyn. of the Geospheres, Moscow, Aug Latter, A. L., R. E. Lelevier, E. A. Martinelli, and W. G. McMillan, A method of concealing underground nuclear explosions, J. Geophys. Res., 66, 2929, Lee, E., M. Finger, and W. Collins, JWL equation of state coefficients for high explosives, Rep. UCID-16189, Lawrence Livemaore Natl. Lab., Livermore, Calif., Jan Murphy, J. R., An evaluation of the factors influencing the seismic detection of decoupled explosions at regional distances, Rep. SSS-R , Areas Control and Disarmament Agency, Washington, D.C., July Murphy, J. R., J. L. Stevens, and N. Rimer, High frequency seismic source characteristics of cavity decoupled underground nuclear explosions, S-CUBED Tech. Rep. SSS-TR , Rep. AFGL-TR , Geophys. Dir., Phillips Lab., Hanscom Air Force Base, Mass., Murphy, J. R., N. Rimer, and J. L. Stevens, Comment on 'seismic decou- pling with chemical and nuclear explosions in salt' by L. Glenn and P. Goldstein, J. Geophys. Res.,,845, Reixfite, R. E., J. A. Leverette, A. A. Martinez, D. Murrell, and C. Joachim, High explosion decoupling experiments in hard rock: The Magdalena tests, paper presented at Nuclear Evasion Testing Symposium, Arms Control and Disarmament Agency, Washington, D.C., Jan , Rimer, N., T. Barker, S. Rogers, J. Stevens, and D. Wilkins, Simulation of seismic signals from partially coupled nuclear explosions in cylindrical tumsels, Rep. DNA-TR , Defense Nuclear Agency, Alexandria, Va., Aug Stevens, J. L., J. R. Murphy, and N. Rimer, Seismic source characteristics of cavity decoupled explosions in salt and tuff, Bull. Seismol. Soc. Am., 81, 1272, 1991a. Stevens, J. L., N. Rimer, J. R. Murphy, and T. G. Barker, Simulation of seismic signals from partially coupled nuclear Explosions in spherical and ellipsoidal cavities, S-CUBED Tech. Rep. SSS-FR , Areas Control and Disarmament Agency, Washington, D.C., 199 lb. V. V. Adushkin and I. O. Kitov, Institute for Dynamics of the Geosph6res, Russian Academy of Sciences, 38 Leninsky Prospect, Korpus 6, Moscow, Russia B. W. Barker and J. R. Murphy, Maxwell Technologies, Inc., Suite 12i2, Sunrise Valley Drive, Reston, VA ( bwb maxwell.com; jrm maxwell.com) N. Rimer, Federal Division Maxwell Technologies, Inc., 8888 Balboa Avenue, San Diego, CA (Received February 13, 1997; revised June 12, 1997; accepted August 28, 1997.)