Experimental study of high energy electron interactions in a superconducting aluminum alloy resonant bar M. Baruccif,g M. Bassanb,c B. Buonomoa G. Cavallarid E. Cocciab,c S. D’Antoniob V. Fafoneb,c C. Ligia L. Lollif,g 9 A. Marinia G. Mazzitellia G. Modestinoa G. Pizzellac,a 0 0 L. Quintieria L. Risegarif,g,1 A. Rocchib F. Rongaa P. Valentee 2 G. Venturaf,g S.M. Vinkoa,2 n a J aINFN (Istituto Nazionale di Fisica Nucleare) Laboratori Nazionali di Frascati, I 9 00044 Frascati, Italy ] bINFN Sezione Roma2, I 00133 Rome, Italy c q cDipartimento di Fisica, Universit`a di Tor Vergata, I 00133 Rome, Italy - r dCERN, CH1211 Geneva, Switzerland g [ eINFN Sezione di Roma1, I 00185 Rome, Italy 1 fINFN Sezione di Firenze, I 00185 Sesto Fiorentino, Florence, Italy v 0 gDipartimento di Fisica, Universit`a di Firenze, I 00185 Sesto Fiorentino, 2 Florence, Italy 2 1 . 1 0 9 Abstract 0 : v Peakamplitudemeasurementsofthefundamentalmodeofoscillationofasuspended i aluminum alloy bar hit by an electron beam show that the amplitude is enhanced X by a factor ∼ 3.5 when the material is in the superconducting state. This result is r a consistentwiththecosmicrayobservationsmadebytheresonantgravitationalwave detectorNAUTILUS,madeofthesamealloy,whenoperatedinthesuperconducting state. A comparison of the experimental data with the predictions of the model describing the underlying physical process is also presented. Key words: Gravitational wave detectors, Aluminum alloy, Superconductivity, Radiation acoustics PACS: 04.80.Nn, 74.70.Ad, 61.82.Bg, 65.60.+a 1 Present address: CSNSM, 91405 Orsay Campus, France 2 Present address: Dept. of Physics, Univ. of Oxford, Oxford OX1 3PU, UK 1 Introduction In a pioneering experiment [1] B.L. Baron and R. Hofstadter measured me- chanical oscillations in piezoelectric disks when penetrating high energy elec- tron beams impinged on the disks. The authors outlined the possibility that cosmic ray events could excite mechanical vibrations in a metallic cylinder at its resonant frequency and that they could represent a background for experiments aimed at the detection of gravitational waves (gw). The gw res- onant detector NAUTILUS, a massive (2.3 t) suspended cylinder made of an aluminum alloy (Al5056) that can be cooled down to the thermodynamic temperature of 0.1 K, has been equipped with a cosmic ray detector to study the interactions due to cosmic rays and to provide a veto against the induced events in the antenna. The results on the cosmic ray observations made by NAUTILUS can be summarized as follows: 1) when the antenna was oper- ated at a temperature T = 0.14 K, well below the transition temperature from normal-conducting (n) to superconducting (s) states of the material, the rate of high energy signals due to cosmic ray showers was larger than the expectations based on the model describing the underlying physical processes [2,3]; 2) there was no evidence of this feature when the antenna was operated at T = 1.5 K, well above the transition temperature [4]. From one side the hypothesis that this behavior was linked to the conducting state of the an- tennaandontheothersidetheincompleteknowledgeatverylowtemperature of the thermophysical and thermodynamic parameters needed by the model have motivated an experiment (RAP) to measure the longitudinal oscillations of suspended cylindrical bars exposed to electron beam pulses of controlled energy and intensity. The experiment, performed at the Beam Test Facility (BTF) [5] of the DAFNE Φ-factory complex in the INFN Frascati Laboratory, has already obtained the following results: 1) the measurements over a wide temperature interval (4.5 K ≤ T ≤ 264 K) on a bar made of the same alu- minum alloy as NAUTILUS have confirmed with good precision the validity of the model [6]; 2) the measurements on a pure niobium bar operated in the n and s state have demonstrated that the oscillation amplitude of the bar induced by the interaction with the beam depends on the state of conduction of the material [7]. In this letter we report on the measurements made on the aluminum alloy bar above and below the temperature of transition between the s and n state. In particular, we present a description of the model (Section 2), a summary of the experimental setup (Section 3), the collected data and analysis (Section 4) and the comparison between the data and the model (Section 5). 2 2 Discussion of the thermo-acoustic effects A pressure pulse is generated in a suspended cylindrical bar in the n state following the interactions of an elementary particle with the bulk. This sonic pulse,duetothelocalthermalexpansioncausedbythewarmingup,relatedto the energy lost by the particle crossing the material, determines the excitation of the vibrational modes of the bar. In the experiment of Ref. [8] an aluminum bar was exposed to a proton beam and the theoretical expectations were based on a model in which the “amplitude of the fundamental longitudinal mode of oscillation”, hereafter referred to as Amplitude, is given by: 2αLW B = (1) 0 πc M V forabeamhittingthecenterofthecylindergeneratrix.Inthepreviousrelation L, M are respectively length, mass of the cylinder, W is the total energy loss of the beam in the bar, α is the linear thermal expansion coefficient and c is V the isochoric specific heat. The ratio of the thermophysical quantities α and c is part of the definition of the Gru¨neisen parameter of the material: V βK T γ = , (2) ρc V where β is the volumetric thermal expansion coefficient (β = 3α for alu- minum), K is the isothermal bulk modulus and ρ is the mass density. The T parameter γ is a very slowly varying function of the temperature when the material is in the n state. Solution (1) is a particular case of a more general treatment of the problem, which includes the paths of the interacting parti- cles in the bulk other than the coordinate of the impact point [9,10,11]. By the introduction of a vector field u(x,t) describing the local displacements from equilibrium, the amplitude of the mode k of the cylinder oscillation is proportional to: ∆Ptherm g therm= A(cid:48)I k k ρ (cid:12) (cid:12) γ (cid:12)dW(cid:12) = (cid:12) (cid:12)I , (3) ρ (cid:12) dx (cid:12) k (cid:12) (cid:12) where ∆Ptherm is the pressure pulse due to the sonic source previously de- scribed, dW/dx is the specific energy loss of the interacting particle, A(cid:48) is the cross section of the tubular zone centered on the particle path in which the effects are generated and I = (cid:82) dl(∇·u (x)) is a line integral over the k k particle path involving the normal mode of oscillation u (x). The Amplitude, k 3 as given by (1), can be obtained starting from (3) for a thin bar (R/L (cid:28) 1, where R is the bar radius) and for particles hitting the central section. In the following, for the material in n state, we will compare the measured values of Amplitude to the expected value: X = B (1+(cid:15)) , (4) therm 0 where (cid:15) is a corrective parameter estimated by a Monte Carlo (MC) simula- tion [6], which takes into account the solutions O[(R/L)2] for the modes of oscillation of a cylinder, the transverse dimension of the beam at the impact point and the trajectories of the secondary particles generated in the bar. The value of (cid:15) for the aluminum alloy bar used in the experiment is estimated by MC to be -0.04. When the material is in the s state, an additional sonic source could be due to the local s-n transitions in zones centered around the interacting particle path [9,10]. The additional contribution to the amplitude of the cylinder oscillation mode k is proportional to: ∆Ptrans g trans= A(cid:48)(cid:48)I k k ρ γ (cid:20) ∆V ∆S(cid:21) = K +γT A(cid:48)(cid:48)I , T k ρ V V where ∆V and ∆S are the differences of the volume and entropy in the two states of conduction, while A(cid:48)(cid:48) is the cross section of the tubular zone cen- tered on the interacting particle path and switched from s to n state, which is given by A(cid:48)(cid:48) = (dW/dx)/(∆H/V) [12,13] involving the difference of en- thalpy, H, among the two states. The differences can be expressed in terms of the thermodynamic critical field H and it follows, in first approxima- c tion, that [14,15]: ∆V/V=(V −V )/V=H (∂H /∂P)/(4π) and ∆S/V=(S − n s c c n S )/V=−H (∂H /∂T)/(4π). Moreover, by using the difference (∆G/V=(G − s c c n G )/V=H2/(8π)) of the Gibbs free energy among the two states and by mak- s c ing the hypothesis that H has the parabolic behavior H (t) = H (0)(1−t2), c c c where t = T/T and T is the transition temperature, it follows that ∆H/V = c c H2(0)(1−t2)(1+3t2)/(8π). In order to compare the observed data with the c model predictions, we will use the ratio R of the contributions to the Am- plitude due to local transition effects (X ) and to thermal effects in the n trans state (X ). R can be expressed as: therm X g trans trans 0 R= = X g therm therm 0 4 (cid:34)K ∆V ∆S(cid:35)(cid:20)∆H(cid:21)−1 T = +T , (5) γ V V V due to the existent proportionality between the mode amplitude and g. In an alternative scenario, which takes into account that local transitions do not occur, the Amplitude expected values are given by the relation (4) making use of the α and c values for the s state. V 3 Experimental setup TheexperimentsetuphasbeenfullydescribedinRef.[6].Herewebrieflyrecall that the test mass is a cylindrical bar (R = 0.091 m, L = 0.5 m, M = 34.1 kg) made of Al5056, the same aluminum alloy (nominal composition 5.2 w% Mg and 0.1 w% of both Cr and Mn) used for NAUTILUS. The bar hangs from the cryostat top by means of a multi-stage suspension system insuring an attenuation on the external mechanical noise of -150 dB in the 1700-6500 Hz frequency window. The frequency of the fundamental longitudinal mode of oscillation of the bar is f = 5413.6 Hz below T = 4 K. The cryostat 0 is equipped with a 3He refrigerator, capable of cooling the bar down to T ∼ 0.5 K. The temperatures are measured inside the cryostat by 10 thermometers controlled by a multi-channel resistance bridge. In particular, a calibrated RuO resistor detects the temperature of one of the bar end faces with an 2 accuracy of 0.01 K for T (cid:46) 4 K. Two piezoelectric ceramics (Pz), electrically connected in parallel, are inserted in a slot cut in the position opposite to the bar suspension point and are squeezed when the bar shrinks. In this Pz arrangement the strain measured at the bar center is proportional to the displacement of the bar end faces. The Pz output is first amplified and then sampled at 100 kHz by an ADC embedded in a VME system, hosting the data acquisition system. The measurement of the Pz conversion factor λ, relating voltage to oscillation amplitude, is accomplished according to a procedure based on the injection in the Pz of a sinusoidal waveform of known amplitude, withfrequencyf andtimedurationlessthanthedecaytimeofthemechanical 0 excitations and on the subsequent measurement of Amplitude. The procedure is correct if R/L (cid:28) 1 and a 6% systematic error in the determination of λ was found. Amplitude is measured according to X = Vmeas/(Gλ), where G 0 is the amplifier gain and Vmeas is the maximum of the signal component at 0 frequency f , which is obtained by Fast Fourier Transform algorithms applied 0 to the digitized Pz signals. The sign of Amplitude is taken positive or negative according to the sign of the first sampling above the noise in the waveform generated by the Pz and sampled by the ADC. BTF delivers to the bar single pulses of ∼ 10 ns duration, containing N electrons of 510 ± 2 MeV energy. e N ranges from about 5 × 107 to 109 and is measured with an accuracy of e 5 ∼ 3% (for N > 5 × 108) by an integrating current transformer placed close e to the beam exit point. MC, already introduced in Section 2, estimates an average energy lost (cid:104)∆E(cid:105)± σ = 195.2±70.6 MeV for a 512 MeV electron ∆E interacting in the bar and, consequently, the total energy loss per beam pulse √ is given by W = N (cid:104)∆E(cid:105) , σ = N σ . e W e ∆E 4 Measurements and data analysis Samples of Al5056 obtained from the same production batch of the test mass have been used to characterize the material at very low temperatures. The measurement of the transition temperature to the s state conducted using the mutual inductance method gives the value T = 0.845±0.002 K and a total c transition width of about 0.1 K. The smaller value of T with respect to pure c Al (1.18 K) could be ascribed to the presence of Mn impurities in the alloy. In fact, experimental studies on AlMn polycrystalline alloys have shown that T c was depressed down to 0.868 K and 0.652 K for Mn concentrations of 440 ppm and 900 ppm, respectively [16]. Moreover, Al50XX alloys contain inclusions of the extremely complex (MgAl) β phase [17] and the characterization of su- perconducting properties of the alloy β−Al Mg shows that T = 0.87 K [18]. 3 2 c Specific heat data for Al5056 are available in literature [19], however, in order to completely characterize the production batch, we have performed c mea- V surements above and below T (Fig. 1) using the calorimetric method of Ref. c [20]. In the temperature interval 0.9 K ≤ T ≤ 1.5 K the fit of the data points, which have an accuracy of 5%, to the function c /T = Γ+BT2 gives the val- V ues Γ = 1157±31 erg cm−3 K−2 for the electronic specific heat coefficient per unit volume in the n state and B = 0.14±0.01 mJ mol−1 K−4 for the lattice contribution. If the superconducting properties of Al5056 can be described by the BCS theory, then H (0) ≈ 2.42 Γ1/2T ≈ 70 Oe. An independent check of c c the c (T) behavior is obtained by the measurements at the bar end face of V the temperature increments due to energy released by each beam pulse. The main features of the c (T) behavior, as obtained by the calorimetry, are well V reproduced by this method (inset of Fig. (1)). The full set of Amplitude measurements (X) normalized to the energy de- posited per beam pulse (W) in the explored temperature interval is shown in Fig. 2. For T ≥ 0.9 K, above T , X has a strict linear dependence on c W, as expected from the relation (1). The linear fit X = bW (Fig. 3) gives b = (2.42±0.17) 10−10 m/J, where the error is determined by the quadrature of the beam monitor (3%) and λ determination (6%) accuracies. The onset at T ∼ 0.9 K and the behavior of the superconducting effects are shown in Fig. 2. As T decreases, the normalized Amplitude becomes negative, indicating that a compression rather than an expansion is generated by the beam interaction in the bulk. Its absolute values is greater than b, the normalized Amplitude 6 Fig. 1. Al5056 specific heat: calorimetric measurements (5% accuracy). The inset shows the calorimetric measurements interpolated by a polynomial (continuous line) and the independent c determination based on the temperature increments at one V end face of the bar (dots). value measured in the n state. The increase in the absolute value of the Am- plitude explains the effects seen in cosmic ray observations by NAUTILUS, when operated at T = 0.14 K, as due to the conduction state of the material. Furthermore, X does not linearly depend on W at fixed T, opposite to what has been observed [7] in pure Nb in the s state. The dependence of X/W on W in the s state is shown in Fig. (4) representing the data in four non- overlapping bands of W. This fact has an impact on the quantification of the enhancement of the absolute value of the Amplitude below and far away from T . Fig. 5 shows the averages of |X/W| and W in four bins of data collected in c the temperature interval ranging from 0.55 to 0.60 K, together with the best fit given by the exponential (cid:104)|X/W|(cid:105) = (8.31±2.88)10−10 e(−26.2±6.3)(cid:104)W(cid:105) m/J. TheaverageenergydepositedbythecosmicraysinteractingintheNAUTILUS 7 Fig. 2. Measured values of Amplitude (X) normalized to the energy (W) deposited in the bar per beam pulse vs. temperature (T). antenna is in the order of 10−8 J [21], much lower than that released in our test mass by the beam pulse. We use the ratio F = (cid:104)|X/W|(cid:105)/b as a factor quantifying the Amplitude enhancement in the s state with respect to the n one. With reference to Fig. (5), we obtain F=3.4±1.2 by the extrapolation of the fitting exponential to (cid:104)W(cid:105) = 10−8 J. A value of F ∼ 3.5 is consistent with the cosmic ray observations made by NAUTILUS in the s state [21]. 5 Comparison with the model The Amplitude (X) linearly depends on the deposited energy (W) in the model described in Section 2, while a X/W dependence on W is observed in thedata.Therefore,wetrytocomparethemodelpredictionstothedatainthe hypothesisthatthelineardependenceofX onW isattainedatverylowvalues of energy deposition. The application of the model for the expected value 8 Fig. 3. T≥ 0.9 K (n state); Measured values of Amplitude (X) vs. the energy de- posited per beam pulse (W). The slope of the fitted line is b = 2.42 10−10 m/J. (X ) computation of the Amplitude in the s state requires the knowledge exp of 1) the thermophysical parameters α and c of the material in order to n V,n evaluate X for the n state below T and 2) the dependence of H on T therm c c and P for calculating X via H and its derivatives ∂H /∂T and ∂H /∂P. trans c c c The use of relations (1), (2), (4) and (5) allow us to write: X X exp therm = (1+R) W W X (cid:40) (cid:20) ∆V ∆S(cid:21)(cid:20)∆H(cid:21)−1(cid:41) therm = 1+ Λ +T (6) W V V V with: 2ρL(1+(cid:15)) Λ = 3πMXtherm W 9 Fig.4.DataofFig.(2)orderedin4bandsofdepositedenergy.Eachbandisidentified by (cid:104)W(cid:105)±σ . A dependence of X/W on W can be seen in the s state, e.g. at T=0.6 W K. The requirement 1) cannot be fulfilled due to the lack of knowledge of α for n Al5056 and we therefore assume that X /W = b also in the temperature therm interval0.5K (cid:46) T ≤ T ,duetothefactthatγ ,inthisinterval,isexpectedto c n have almost the same value as that assumed at slightly higher temperatures. In relation to requirement 2), we derive ∂H /∂T at T < T from the H c c c parabolicdependenceont,assumingthattheunknowndependenceof∂H /∂P c on t at P = 0 for Al5056 is equal to that of pure Al. Under this hypothesis, ∂H /∂P can be deduced by interpolating the tabulated values of H as a c c function of T and P contained in Ref. [22]. Inserting numerical values in relation(6)givesanaverageof(cid:104)X /W(cid:105) = (−18±1)10−10 m/Jintheinterval exp 0.55 ≤ T ≤ 0.6 K, where the error does not include systematic contributions deriving from the assumptions made. This is to be compared to (cid:104)X/W(cid:105) = (−8.3 ± 2.8)10−10 m/J, obtained at W = 10−8 J from the measurements in the same temperature range (see Fig.(5)). This discrepancy can be ascribed to the fact that the model, as mentioned by the authors of Ref. [10], considers 10