2.1. Pulse Width Modulated Deflection System
The concept behind this method is to use an electrostatic deflector as a beam chopper. A beam chopper is a device that periodically interrupts the particle beam. In our microbeam line, it must travel through the set of aligned object and opening slits separated by a distance of approximately 6 m. Normally, to turn off the beam, a voltage pulse that deflects the beam outside the aperture slits is applied. In this case, the deflector is employed as a “beam blanker”. In the current operation, it is used with the parallel deflecting plates at high potential. That is, the beam is normally off and the pulse beam is generated when it falls into the slit aperture at the bottom of an excitation voltage pulse, annulling the continuous voltage applied to the plates.
With this technique, a continuous particle beam is vertically deflected over the apertures of the slits and, at a given frequency, the average beam current arriving at the device is reduced proportionally to the duty cycle of the deflecting signal. Therefore, if a given frequency has a typical period f
, this period (t
) will be made up of two components: the period when the high voltage is applied and the beam suppressed (thigh
), and the period when the voltage is close to 0 V and the beam is passing up to the sample (tlow
). Then, the duty cycle is defined as follows:
Two different regimes can be distinguished. If the temporal distance between two consecutive ions is, on average, smaller than tlow
the resulting beam is going to be composed of bunches. Each bunch is going to be made up of several ions depending on the input current. Section 2.7
shows an application for this mode of operation. In the case that the temporal distance between consecutive ions is greater than tlow
, a quasi-continuous beam is generated at the input of the aperture slits. This means that each ion is going to be separated from the next for a period larger than thigh
. The final beam current will depend (on average) on the duty cycle only. However, the regime (continuous or pulsed) will also depend on the frequency.
Physically, the deflector is placed inside the vacuum tube immediately after the object slits. The beam deflector consists of two 15 cm long parallel metal sheets with a 2 mm gap. The conducting layers are made up of thin copper foils laminated onto a non-conductive substrate using epoxy printed circuit board technology. The deflector system is schematically presented in Figure 1
For an ion beam of energy EP
and charge qp
, the minimum voltage required across the deflecting plates to ensure that the beam diameter (BD
) at the aperture slits is removed completely from the aperture’s diameter DAS
is given by:
is the electron charge and the rest of the symbols are as defined in Figure 1
. Therefore, LDP
should be taken as large as possible and dDP
as small as possible to minimize the required voltage. In practice, LDP
= 15 cm and dDP
= 2 mm were chosen to minimize the capacitance between the plates, and also to achieve a fast switching of the beam. This plate configuration resulted in a capacitance of ≈10 pF. In our beamline, the deflector was situated 30 cm after the object slits and 5.8 m before the aperture slits.
As it follows from this formula, the necessary voltage to completely deflect the beam out of the opening slits depends on the beam diameter, which in turn depends on the emittance of the accelerated ion beam, that is, the beam quality repeatability. In each case, there will be a necessary minimum potential (VDPmin) to deflect the beam. Usually, in our old accelerator, to reduce the current of the beam, deflection potentials are quickly tested and, to minimize the deflection time, a slightly higher voltage than the necessary minimum is used. For 16O5+ beam at 50 MeV energy with BD = 4.6 mm diameter at the anti-scattering slits with DAS = 0.75 mm a minimum deflection voltage of VDPmin ≈ 70 V is necessary to blank the beam.
The deflection signal is produced by an arbitrary waveform generator (Rigol® DG4162, Rigol Technologies Inc., Beijing, China). The output signal feeds a high voltage amplifier (Technisches Büro S. Fischer, Ober-Ramstadt, Germany) connected to the electrostatic deflection plates in order to turn on the beam.
2.2. Beam Current Measurement
Simultaneously with the application of the system to change the beam current, one of the most important parameters to consider in any system that is dedicated to the study of radiation damage is the control and measurement of the dose.
The Pelletron charging system of our accelerator causes instability in the beam energy producing fast fluctuations in the current [14
], which depends on the stability of the ion source, focusing optics and beam transport, and it may cause slower variations if the operation is not optimal. In order to have an appropriate dose quantification, a precise measurement of the beam current arriving at the sample is required. The most usual method for measuring current for PIXE experiments involves a Faraday cup (FC) placed behind the sample.
State-of-the-art FCs can measure currents as low as 10 fA [15
]. For a microbeam of 16
with a spot cross-section of a few square microns, 10 fA is equivalent to 104
ions/s, which represents a high current for SEE studies. Electronic devices are too thick for the beam to pass through, thus making it necessary to find other ways of dose normalization (online current measurement). Since in SEE studies the current is usually formed by tens or hundreds of ions per second and must be monitored continuously, the use of a FC to measure the current is discarded. Other indirect continuous measurement techniques such as Rutherford backscattering spectrometry (RBS) or X-ray production (PIXE) in the metallic coverage of the devices are only applicable for high-efficiency detectors and larger currents and/or doses.
On the other hand, direct measurement of the particle current as in Scanning transmission ion microscopy (STIM) requires that the samples be completely traversed by the primary ions.
Considering the fact that in SEE studies in electronic components the current cannot be measured continuously, we decided to use a periodical monitoring method, that is, measure several times during the irradiation interrupting the beam without modifying the irradiation conditions. This can be done by placing a particle detector that intercepts the beam path at regular intervals, fast enough to avoid significant dead times between the measurement and the irradiation periods. In each experiment, it is possible to select the periodicity and time of the measurements to obtain an optimum measurement of the particle beam that reaches the sample.
A similar measurement system was described in [16
]. In our case, the detector is placed inside the measuring chamber, approximately 10 cm before the sample. Since SEE experiments are usually performed with heavy ion beams with tens of MeV’s energy, producing irreversible damage in the semiconductors, it is necessary to use a cheap and easy-to-replace detector as a low-cost alternative to surface barrier detector. As in [16
], we decided to use a Hamamatsu S1223-01 Si-PIN photodiode (Hammamatsu, Shizuoka, Japan). This photodiode is widely used as an ion detector and there is extensive literature on its good radiation hardness [17
]. Thus, the installed setup utilizes a fast electrostatic beam deflector to reduce the current and a movable Si PIN-diode directly measure the beam intensity. This method was used to test SEE produced in a PLL implemented on CMOS 90 nm technology [3
2.3. Voltage Measurement
To attenuate the beam current, we applied a square signal using different pulse lengths (duty cycles) at a given frequency. This way, one of the plates of the deflector is connected to ground and the other to the variable potential. Thus, ions will reach the sample while the potential lies between 0 and VDP
volts. The output pulses VDP
) of the deflecting plates are shown in Figure 2
= 70 V represents the voltage below which a 16
beam at 50 MeV starts moving towards to the target until it arrives to it. The time intervals indicated at the top of the graph represent the effective enable temporary windows for the shortest pulses allowed by the fast high voltage (HV) amplifier, that is, the times during which the VDP
voltage lies below 70 V. The duty cycle in each case will also depend on the pulse frequency.
In Figure 2
the negative overshoots and positive undershoots of the shortest signals are generated by the fast HV amplifier.
The minimum rise time of 220 V deflection pulse provided by the high voltage amplifier is 3.5 ns/V. Assuming a uniform beam of 16O5+ at 50 MeV with a current of approximately 3.2 pA, the temporal distance between particles is 0.25 μs. For this beam, according to the calculation above, the temporary enable window VDP occurs between 70 V and 0 V. For a pulse frequency of 1 kHz and an effective duty cycle of 0.025% the temporary enable window between 70 V and 0 V is 253 ns long.
Under these conditions, ion passage is allowed 1000 times/s and the object and collimator slit openings remain unchanged. This makes the operation of low current techniques, such as STIM and IBIC very convenient to use. There are commercial HV switches that can produce high voltage with a few tens of ns rise time. In [22
] a beam deflector uses fast high voltage push-pull MOSFET switch (Behlke Electronics Gmbh, Kronberg, Germany) having 40 ns rise time for 1 kV pulses. The use of a device with these characteristics would allow us to significantly improve the current reduction range.
2.5. Particle Arrival Statistics
For experiments where a low cumulative dose is desired, it is not necessary to analyze the time distribution of the ions. However, to ensure that SEE happen, it is required that the average time interval between the arrivals of two consecutive ions be much greater than the characteristic response time of the device under study. For MeV ions impinging on electronic components, the pulse duration can range from tens of ps [23
] to tens of µs [24
]. However, the main factor limiting the number of pulses that can be recorded is usually the number of waveforms per unit of time that the oscilloscope used for the SEE detection can process. This number of waveforms can be in the order of a few thousands per second or less.
Accordingly, we measured the number of ions in 1 ms intervals during one minute with a MultiChannel Scaler (Ametek, Oak Ridge, TN, USA) and a silicon PIN diode as a detector. The particle beam intensity without deflection was ≈ 22,000 ions/s and the deflector was configured using a 0.1% duty cycle and 100 Hz frequency.
In order to study the arrival time distribution, we made a histogram showing the time between two consecutive ions hitting the target. As it is shown in Figure 4
, the resulting histogram can be approximated by an exponential decay function with a mean time of 45 ms. It is well known that Poisson processes show this kind of distribution when the time between events is analyzed. As a result, we can assume that the process is random and the mean time is consistent with the expected value (for a quasi-continuous beam). If the beam were pulsed instead of quasi-continuous, a large number of events should be observed on the first time interval.
In order to measure the beam current when a thick sample is placed on the target, we designed a movable detector that is periodically placed in front of the beam, inside the vacuum chamber. The interposition of the detector is performed by using a servo-motor handled by an Arduino microcontroller based system. This system measures the number of incident ions at regular intervals with a transit time (dead time) of about 60 ms/measurement. To verify that the measured current with this method is representative of the number of ions impinging on the target, another PIN diode was placed as target and the counting for both detectors was recorded. The servo-motor was programmed to enable the beam transmission at intervals of 1 s. Figure 5
shows the measured current on the servo PIN-diode. Inverted peaks correspond to time intervals where the servo PIN-diode is out of the beam. The circles represent the counting rate of the target PIN-diode.
shows the histogram of the counting rate at the target PIN-diode. The width of the distribution (about 10%) is principally given by fluctuations of the beam current, and it is the principal source of uncertainty in dose determination. For this particular beam conditions (about 103
ions/s), the difference between the averages counting rates measured on the mobile PIN-diode and on the target PIN-diode is about 0.35σ. The maximum current that can be achieved with this technique is limited by the time response of the detector and amplifier system which is in the order of 103
ions per second for most of the solid-state ion detectors (besides the damage caused by heavy ions with tens of MeVs energy). With this technique, we can obtain total accumulated ions in the range of 103
ions for a low current and short time irradiation up to 3.6 × 106
ions for a one-hour long irradiation with a relative uncertainty of ≈10% covering an area of 10 μm2
2.6. Low Current Application: SEE in a CMOS Output Buffer
As an example for SEE measurements with the previously described system, two digital output buffers (0.5 µm CMOS (Complementary Metal-Oxide Semiconductor) process) were tested. As shown in Figure 7
the device was formed by cascading four CMOS inverters with increasing sizes from input to output. This kind of circuit is used to increase the driving capability of a logic gate made by minimum size transistors for large load capacities, so it is present in the output of most digital integrated circuits (ICs). Each inverter is composed of one PMOS (p-type MOS) and one NMOS (n-type MOS) so, under normal conditions, only one of the transistors is ON (saturation) while the complementary is OFF (cut-off).
To test both operation states (high state follower and low state follower), each buffer had its input fixed to one of the logic levels: BUFFER 0 had its input fixed at ground potential (P1/N2/P3/N4 in the ON state) while BUFFER 3 had its input at VDD (drain to drain voltage) (5 V) potential (N1/P2/N3/P4 in the ON state).
A 3 µm beam spot of 32
at 75 MeV was used for the experiment. The IC was mounted on a printed circuit board inside the high vacuum microbeam chamber. The outputs of both buffers were connected outside the chamber through BNC connectors and acquired by a 1.5 GHz bandwidth and 20 GS/s oscilloscope (Teledyne Lecroy WavePro 715Zi, Teledyne Lecroy, Chesnut Ridge, NY, USA) with the trigger level slightly above the noise. The input signal of the XY deflector coils amplifier was acquired in two channels of the oscilloscope. Two different transients are shown in Figure 8
As expected, the measured output pulses were negative for BUFFER 3 (High state follower) and positive for BUFFER 0 (Low state follower).
shows the X-Y map of event triggers overlapped with the actual layout of the tested buffers. The SEE coordinates were over the transistors N1/P2/N3/P4 of BUFFER 3 and over P1/N2/P3/N4 of BUFFER 0. All of these devices are the ones that were in cut off operation mode.
2.7. Radiation Damage in LiNbO3 Applying the Constant Gradient Technique
In order to show the potential of using the electrostatic deflector on demand as beam current controller, it was used in combination with the microbeam scanning system with the purpose of generating a smooth damage gradient with different applied doses on a piece of LiNbO3
x-cut. Previous studies [1
] show that using fluences of 5 × 1012
at 70 MeV generate lattice damage in a way that the etching rate in an HF (50%) acid increases drastically. We used a similar projectile as in [4
], a 5 μm beam spot of 32
at 75 MeV energy and 250 × 250 µm2
scanned area. The electrostatic deflector was used to reduce the flux using a pulse width modulation (PWM) signal phased with a triangular (sawtooth) signal.
The used waveform generator allows to produce signals of up to 16,384 points. We divided the data string into 128 segments of 128 points each. Every segment was set up in order to have a duty cycle of ≈0.78% larger than the previous one. In the second channel, a constant voltage that increases every 128 points (stepped waveform) was configured (Figure 10
). That signal was connected to one of the inputs of the scanning amplifier corresponding to the X-axis. The Y-axis input was connected to another triangle wave generator, which scans the beam along the Y-axis during the 128 steps intervals for the different X positions. The frequencies were 256 Hz for the X-axis and 2 Hz for the Y-axis. The fluence was controlled measuring the X-rays emitted by the sample during the irradiation (Figure 11
). We had previously normalized the X-ray production of the sample with the ion current. At the end of the experiment, the fluence was linearly decreasing from 8.08 × 1013
to 6.31 × 1011
in 128 steps.
Furthermore, in order to characterize the damage in the material, micro-Raman technique was used. The spectra were recorded using a LabRAM HR Raman system (Horiba Jobin Yvon, Kyoto, Japan), equipped with two monochromator gratings and a charge-coupled device detector. An 1800 g/mm grating and a 100 μm hole results in a spectral resolution of 1.5 cm−1. Polarized light of the 514.5 nm Ar laser line was used as the excitation source. Measurements were taken in backscattering geometry, with a 50× magnification under a confocal microscope coupled to the spectrograph. Under these conditions, the spatial resolution of the system is about 5 μm. In this range, the irradiated zone can be easily distinguished using conventional microscopy due to the changes in the optical properties of the material. The spectra were taken from the edge of the irradiated zone up to 280 μm further. The 368 cm−1 peak area was used to normalize the spectra because it showed slight or no change at different sample positions.
As it can be seen in Figure 12
, as the doses increases, peaks present in the unirradiated horizontal polarization (Y(XX)Y) gradually appear in the irradiated vertical polarization (Y(XZ)Y), and vice versa. In order to analyze this behavior, the sample was metalized with 40 nm gold and the irradiated zone was measured using an optical profilometer. As shown in Figure 13
there is a gradual increment of the sample thickness caused by the accumulated dose (swelling).
presents photo-elastic properties [25
] so mechanical stress can induce a shift in the polarization angle of the light that passes through the crystal. Probably, irradiation is introducing mechanical stress in the crystal. This effect may be the reason why Raman peaks present on each polarization are gradually appearing on the other, as irradiation dose increases.
Although it is not the purpose of this paper to explain the causes of these changes, it is clear that this technique allows characterizing the evolution of damage in a given material for many different doses making a single irradiation.