Miniature silicon Michelson interferometer characterization for dimensional metrology

Sensors and Actuators A 223 (2015) 141–150

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna

Miniature silicon Michelson interferometer characterization for dimensional metrology Hichem Nouira a,∗ , Jean-Pierre Wallerand a , Maurine Malak b , Anne-Franc¸oise Obaton a , José Salgado a , Tarik Bourouina c a

Laboratoire Commun de Métrologie (LNE-CNAM), Laboratoire National de Métrologie et d’Essais (LNE), 1 Rue Gaston Boissier, 75015 Paris, France Ecole Polytechnique Fédérale de Lausanne, Rue de la Maladière 71B, CH-2002 Neuchâtel, Switzerland c Université Paris-Est, laboratoire ESYCOM, ESIEE-Paris, Cité Descartes, 2 Bd Blaise Pascal, 93162 Noisy-Le-Grand Cedex, France b

a r t i c l e

i n f o

Article history: Received 20 May 2014 Received in revised form 24 December 2014 Accepted 29 December 2014 Available online 7 January 2015 Keywords: MEMS (Micro Electro Mechanical Systems) Optical micro-probe Michelson micro-instrument Dimensional metrology Error sources Uncertainty evaluation

a b s t r a c t Dimensional metrology applications require performing measurements of profile and form/shape of parts at a nanometer-level of accuracy. Therefore, the metrological characteristics of a miniature optical micro-probe based on a Michelson interferometer are evaluated. The optical micro-probe is designed to perform dimensional measurements with a target uncertainty in the order of few nanometers. Two microprobe designs having reflection-transmission ratios of 75–25% and 25–75%, are characterized. Two optical setups have been implemented as well: firstly, using a single laser diode with a 1550.3 nm wavelength and secondly using a tunable laser source in the C-L bands. The characterization of the two micro-probes is performed using a new ultra-precise test bench, with respect to both dissociated metrology structure and Abbe principles. The experiments allow the evaluation of the error sources such as: stability, axial motion errors (residual errors), material dependence, tilt angle and roughness of the tested object. The experimental results revealed that dimensional measurements could be achieved with nanometer-scale errors, ranging from 2 nm to 15 nm, depending on the probe design and the reflectance of the device under test. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Recent technological advancements led to an increasing demand for precise measurements of dimensions, topographic profiles and shapes of smaller components and tinier features. The components can vary from fuel injection nozzles, micro-lenses, fluidic channels and micro-piston-cylinders for pressure metrology to features and structures on MEMS chips. The progress in miniaturization puts stringent requirements on quality assurance in-line with manufacturing process. Therefore, numerous tactile single scanning micro-probes have been designed by National Metrology Institutes (NMIs), but most of them have quite similar designs, resolution and operation principle [1–6]. For instance, a white point optical stylus system has been developed by the Bosch Company and exclusively integrated on the Mar-Form-MFU100 apparatus [7]. It consists of a millimeter-sized optical stylus probe for form and shape measurements with less than 1 ␮m accuracy [8]. Its operation principle is based on a short coherence

∗ Corresponding author. Tel.: +33 1 40 43 37 63. E-mail address: [email protected] (H. Nouira). http://dx.doi.org/10.1016/j.sna.2014.12.031 0924-4247/© 2015 Elsevier B.V. All rights reserved.

heterodyne interferometry divided into two subsystems: a modulation interferometer and a small robust optical probe connected through a single monomode optical fiber. The interferometer employs a broadband fiber-optic light source and works at the optical wavelength () of 1550 nm and the synthetic wavelength of 36 ␮m. This paper deals simultaneously with the characterization of two novel interferometric optical micro-probes and the investigation of their best deployment. A new high precision test bench is designed to ensure measurements with nanometer level of accuracy. The characterization of the micro-probes was carried out statically (in a stepwise operation mode) and compared to the data acquired from RENISHAW laser interferometer directly traceable to International System of Units (SI) meter definition [9]. The impact of different materials (aluminum, steel, silicon, copper, gold and ceramic) on the behavior of the optical micro-probes was evaluated. Those target surfaces are commonly used in dimensional metrology and in other sectors. Moreover, the effects of both angular misalignment and roughness of the artifact on the behavior of the micro-probes have been investigated. Optical tests have been performed with a first optical setup based on the use of a laser diode source working at the wavelength of 1550 nm.

142

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

Fig. 1. Photograph of the micro-machined optical interferometer micro-probe together with SEM photography zooming on the Michelson interferometer integrated at its end. The different elements are described along with the transmission and the reflection ports. Magenta arrows designate the injected light, red arrows designate the path of the light beam reflected from the movable mirror (the movable mirror is not present in this figure) and cyan arrows designate the path of the light beam reflected from the reference mirror.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

Following the obtained results, the most accurate micro-probe has been incorporated in a second optical setup where the laser diode source is replaced by a laser source tunable in the C–L bands. Similar tests were performed and the results reveal that the micro-probe can be used in the wavelength band from 1450 nm to 1600 nm. 2. Design of the interferometric micro-probe The optical micro-probe consists of monolithic silicon block (Fig. 1) with a long cantilever beam acting as a probe whose dimensions are 390 ␮m thick, 550 ␮m width, 4 mm length; it involves a relatively large supporting area of nearly one square centimeter used for further handling and assembly to the measurement setup. A Michelson interferometer is integrated at the far end of the cantilever beam. In the interferometer design, the movable test mirror has been removed and it has been replaced by an external reflector (test object or artifact). The interferometer design involves a Beam-Splitter (BS) used to split light between the reference and movable mirrors. The reference mirror is a distributed Bragg reflector consisting of an alternation of four Silicon-Air layers to get a high-reflectance mirror. Two different designs have been presented [10,11]: design “A” and design “B” corresponding to thicknesses of 3.58 ␮m and 3.62 ␮m for the BS, and, to reflectance–transmittance ratios of 75–25% and 25–75% for micro-probes “A” and “B”, respectively. For all Bragg mirrors, the silicon layer thickness is 3.67 ␮m while the air layer thickness is 3.49 ␮m. Photo of the micro-fabricated device is shown in Fig. 1 together with an annotated SEM photo that highlights the various components of the interferometric micro-probes. The micro-fabrication process [12] of the optical micro-probes is realized in ESIEE cleanroom facility.

configurations, the Michelson interferometer is used in the reflection mode without need for any additional output fiber. A single lensed fiber coupled to a fiber circulator was used for both light injection and detection. The lensed fiber has a spot size of 50 ␮m and working distance of 1 mm. National Instruments acquisition card is used to control the movements and to acquire the data screened by the optical detector, connected to the reflection port of the circulator. In the first optical setup, the light beam of the laser diode (ThoroLabs: LM14S2-1550.3 nm) passes successively through the circulator and the input optical fiber (AR-coated lensed fiber) packaged in the input U-groove of the silicon micro-probe. Then, it is reflected from the test object back towards the interferometer (Fig. 2). The reflected interferometric signal is transmitted again through the input U-groove and passes successively through the same input optical fiber (AR-coated lensed fiber), the circulator and

3. Optical setup for characterization of the interferometric micro-probes Two different optical setups have been proposed for characterizing the interferometric micro-probes described earlier. In both

Fig. 2. Optical setup involving the miniature interferometer micro-probe, the laser diode source and the detector. The artifact can be a tiny bore hole or any flat surface.

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

Fig. 3. Schematic of a classical test bench used for the calibration of the probing system. The calibration exactness of the micro-probe depends on the quality of motion of the movable table, the alignment of the micro-probe on the artifact, the environment and the calibration of the physical step-gauge standard. The comparison of the high steps values detected by the micro-probe to the calibrated standard allows calibration of the micro-probe. The calibration of the standard (step gauge) can be achieved on Coordinate Measuring Machine (CMM) tracable to SI meter definition.

the detector (light-wave laser diode controller–ThorLabs: PDA50BEC 800–1800 nm). Only the reflected signal is measured by the detector and then recorded using the acquisition card. The second optical setup consists of a laser source tunable in the C- and L- band from 1465 nm to 1625 nm (Photonetics TUNICS-PRI Tunable Laser Diode Source) connected to the AR-coated lensed fiber through a circulator. The measurement technique is based on wavelength sweep to record the spectral interferogram at the reflection port. The resulting spectrum is post-processed by Fast Fourier Transform (FFT) to determine the optical path difference [10–12].

143

errors (one positioning error Exx and two straightness errors Exy and Exz ) and three rotational errors (roll Ex Rx , pitch Ex Ry and yaw Ex Rz ). For high-precision mechanical guiding systems (namely, linear stages), residual translation errors can attain sub-micrometers values while the residual rotational errors can attain few hundred of micro-radians. As a consequence, the deployment of such method limits the calibration accuracy of the micro-probe to the micrometer level since it cumulates different error sources: motion errors of the micro-probe generated by the mechanical guiding system, calibration uncertainty of the step gauge standard, alignment of the standard along the vertical axis (center and tilt), alignment of the micro-probe with respect to the standard, thermal error, and positioning error. In brief, although this first method is adopted by many research groups, it is not suitable for the calibration of the micro-probe, in particular when looking for a nanometer level of uncertainty, but it is sufficient for demonstrating a micrometer level of accuracy. Based on this analysis, performance tests on the two optical micro-probes “A” and “B” should be undertaken with a method enabling a nanometer level of accuracy. It consists on comparing the data of the tested optical micro-probe directly to those obtained by RENISHAW laser interferometer, (having sub-nanometer resolution) when aligned to the same test object (such as mirror/reflector) according to the Abbe principle [14] (Fig. 4). For this configuration, translation of the test object is detected simultaneously by the laser interferometer and the micro-probe in opposite directions as illustrated in Fig. 4. The test object involves both the Invar block (9) and the three Zerodur reflectors/mirrors. The advantage of such method is that the sources of errors are minimized. Additionally, the micro-probe calibration becomes directly traceable to the International System of Units (SI) meter definition thanks to the practical realization of experiments involving a primary laser (I2 -stabilized He-Ne laser) [15,16]. Consequently, the calibration can be achieved by generating a motorized very small step motion (in the order of 5 nm) over the entire travel range of the micro-probe. In this case, the calibration is never influenced by the motion quality of the mechanical guiding system translating the test object, on which both the micro-probe and the subnanometer laser interferometer are mounted in opposite directions (Fig. 4).

4. High precision mechanical setup

4.1. Design of the mechanical test bench

The characterization of the micro-probes can be carried-out using physical standards with high surface quality such as the VLSI standards [13]. This standard surface is assimilated to a step-gauge made of silicon. Indeed, it can be made of any other kind of material such as: Zerodur, Invar, super-Invar or Stainless steel. The standard should be firstly calibrated on an ultra-high precision coordinate measuring machine [14] with an uncertainty of few tenths of nanometers. Afterwards, the micro-probe is fixed on a mechanical test bench, enabling horizontal motion and then, aligned on the calibrated standard taken as reference element (Fig. 3). The horizontal translation of the micro-probe along the x-axis, as illustrated in Fig. 3, permits the detection of the high steps constituting the standard (step-gauge). Comparison of the detected high steps values to the calibrated standard allows calibration of the micro-probe. Nonetheless, the calibration is usually realized on few steps, which limits the travel range of the micro-probe to a small bandwidth. Additionally the calibration carried out on few steps restrains the comparison to few values only, imparting insufficient data about the micro-probe behavior. Such calibration of the micro-probe is typically realized by a high precision apparatus integrating ball bearing linear stages. Since the shapes and forms of the balls and the linear guide surface are never ideal, any introduced translation (i.e. along the x-axis) provokes residual motion errors: three translation

Performance tests of the interferometric micro-probes “A” and “B” are executed using a new test bench designed to ensure that measurements exhibit a nanometer level of accuracy. The test bench, schematized in Fig. 4, enables two independent motions for the test object along x- and y-axis. Along the x-axis, two independent mechanical translations Tx1 and Tx2 are allowed for coarse and fine displacements. The first translation can be obtained by a sub-micrometric guiding system over a travel range of 63 mm, especially used to adjust the gap between the micro-probe and the test object manually. The second translational guiding system incorporates a piezoelectric actuator with nanometer resolution and a mechanical structure based on four flexible blades to prevent the hysteresis and the stick-slip phenomenon. Therefore a specific manufacturing process was adopted for the flexible blades to guarantee a uniform depth for the flexible zone, which allows a homogenous elastic deformation of this zone. The proposed system generates quasi-perfect nanometer motion in steps of 5 nm over an entire travel range of 90 ␮m. Along the y-axis a mechanical guiding system provides translation Ty with an entire travel range of 50 mm and a sub-micrometric motion step of 0.7 ␮m. A manual rotary stage based on flexible blades provides tilting (Ry ) of the micro-probe around the y-axis, which is helpful for the alignment of the micro-probe with respect to the test object. An additional

144

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

Fig. 4. Schematic of the test bench (the working range is considered as fixed distance necessary for the positioning and the operation of the micro-probe). The test bench insures the comparison of the micro-probe to the laser interferometer data in the same Abbe axis. Then both the micro-probe and the laser interferometer are focused on the same object. Here, the object contains the reflectors and invar block (9). The motion of the movable block along the x-axis is detected by the micro-probe and the laser interferometer but in opposite direction.

manual rotary stage provides the rotation (Rz ) of the test object around the z-axis and around the K point, which corresponds to the intersection of the first and second Abbe axes (Fig. 4). Therefore, the first Abbe axis is mingled with the RENISHAW interferometer laser beam and the optical micro-probe axis when oriented perfectly horizontal. The two sub-nanometer laser interferometers, considered as reference elements, are aligned to two independent movable Zerodur plane reflectors having high surface quality and a flatness of /20, as shown in Figs. 4 and 5. The measured relative displacements Txi (i = 1, 2) and Ty of the block (6) depend on the accumulated phase change ϕ. Applying the formulas in Eq. (1)

and Eq. (2) [17] and counting the  number of fringes k, the measured relative displacement L = Lf − Li can be determined. It is worth to mention that relative displacement depends on the dead path of the interferometer P = Li − LR , the laser wavelength in vacuum 0 and the refractive index of air nair which, in turn, depends on the temperature T, the hygrometry RH and the pressure P [18].

 L = ϕ

(f ) 4



 + (Li − LR )

(f −i) i

 (1)

Fig. 5. (a) Design of the test bench. The metrology loop of the setup passes only through Invar frame (block and rods) and is dissociated from the base plate (1) by flexible blades, (b) photography of the micro-probe fixed in the bench test. (1) Base plate made with Aluminum material, (2) Manuel linear stage: Tx1 , (3) High precision linear piezoelectric stage: Tx2 , (4) Motorized linear stage: Tx1 , (5) Aluminum block, (6) Laser interferometer (2): y-axis, (7) Zerodur reflectors/mirrors, (8) Manuel rotary stage (Rz ), (9) Invar block, (10) Silicon optical micro-probe, (11) Manuel linear stages, (12) Manuel rotary stage based on the use of flexible blade (Ry ), (13) Aluminum block, (14) Invar rods, (15) Laser beams, (16) Laser interferometer (1): x-axis, (17) Aluminum block, (18) Flexible blades dissociating the metrology frame from the supporting holder, (19) Optical fiber for the micro-probe, (20) Renishaw optical fiber.

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150



Evolution of the micro-probe ("A") data Versus the laser interferometer data

(2)

The experiments have been performed in the LNE cleanroom, where both temperature and relative humidity are controlled at 20 ± 0.3 ◦ C and 50 ± 5%RH respectively. The experimental setup is also enclosed in an aluminum shell to improve the homogeneous distribution of temperature around the bench. Moreover, the setup has been mounted on an anti-vibration table to attenuate low frequency vibrations. All motions pilots are controlled using a developed interface in LabView. A systematic and synchronous record of the laser interferometers output, the optical micro-probe output, temperature, hygrometry and pressure data are carried out during the test.

Data (V)

(a)

Data (nm)

0.4

200

0.3

150

0.2

100

0.1

50

0

0

-0.1

-50

-0.2

-100

-0.3

-150

Distance (nm)

nair (i − f ) 0 k + P 4nair (f ) nair (f )

Analog output (V)



L =

145

-0.4 -200 0.15 0.24 0.32 0.41 0.49 0.58 0.67 0.75 0.84 0.93

Laser interferometer data (µm)

Evolution of the micro-probe ("B") data Versus the laser interferometer data

◦ Expansion of the material constituting the metrology loop. This error is estimated to be 2.17 nm. ◦ Variation of the wavelength of the laser interferometer according to the variation in: temperature (T), humidity (RH), pressure (P), dead path (P) and refractive index of air (nair ) in vacuum (v ). For a working distance L of 90 ␮m, the laser interferometer error is estimated to be 2.1 nm. ◦ Cosine error of the measuring entities. This error can be estimated by quantifying the straightness of the motion and the acceptance angle of the laser interferometer. This error is negligible in our case (1 nm). ◦ Resolution of the laser interferometer which is limited by the performance of the Heidenhain counting card IK220. The electronic card accepts 12 bits interpolation. Considering the 4-ways laser instrument, this resolution equates to 0.04 nm. ◦ Abbe error, which is related to the offset and the small angle deviation between the measurand and the axis of measurement. This error is negligible in our case (1 nm). To sum up, the uncertainty budget of the test bench is estimated to few nanometers (less than 5 nm) when considering all the error sources mentioned above.

Data (nm)

(b) 0.25

250

Analog output (V)

Data (V)

The test bench is designed to calibrate the developed microprobe at different measuring ranges and it is based on the application of the dissociated metrology technique (DMT) [19,20]. In order to apply the DMT principle, the Aluminum block (17) supporting the laser interferometer (16) was dissociated from the test bench base made of aluminum, by two flexible blades (Fig. 4). Two Invar rods (14) have been added to rigidly set the measuring entities (Aluminum block (17) and Aluminum block (13)) and to shorten the primary metrology loop. The primary metrology loop represents all physical parts along which the information pertaining to the measurement paths. Consequently, any thermal expansion and/or mechanical deformation of the bench base will not influence the primary metrology loop. Since Invar block (9), supporting the test object and the Zerodur reflectors, is also made of Invar, the metrology frame becomes less sensitive to thermal drift. The test bench presents a second metrology loop passing through the bench base, Aluminum blocks (5) and (13), the optical micro-probe (10) and the laser interferometer (6) acting in the y-direction. Any disturbance of the metrology frame degrades the exactness of the measurement. For this reason, evaluating the performance of the test bench is required. Two kinds of errors can be distinguished: systematic errors (or corrigible errors) and random errors (or noncorrigible errors). Therefore the main error sources are:

0.15

150

0.05

50

-0.05

-50

-0.15

-150

Distance (nm)

4.2. Metrology analysis of the test bench and evaluation of its performance

-0.25 -250 0.19 0.28 0.37 0.45 0.54 0.63 0.71 0.80 0.89 0.97

Laser interferometer data (µm) Fig. 6. (a) Evolution of the micro-probe (“A”) data when changing the gap between the micro-probe and the Zerodur reflector: zoom on the first undulation, (b) evolution of the micro-probe (“B”) data when changing the gap between the micro-probe and the Zerodur reflector: zoom on the first undulation.

5. Experimental results 5.1. First optical setup for micro-probes “A” and “B” The first tests are performed with the optical setup based on the laser diode of fixed wavelength (Fig. 2). A step-by-step axial motion is generated using the piezoelectric stage which provokes the change of the gap between the micro-probe and the test object. The working range between the micro-probe and the test object is kept unchanged at 600 ␮m. Sub-micrometric step motions are generated to cover the measurement range of 1600 nm. An analog sinusoidal output signal is recorded and converted to distance as shown in Fig. 6(a). For micro-probe “B”, conversion of the analog output signal to distance yields a perfect triangular signal as shown in Fig. 6(b), but the analog output of the micro-probe “A” involves noise. Stabilities of the two micro-probes are investigated over a time interval of 30 min and the corresponding results are shown in (Fig. 7). The recorded signals exhibit a Gaussian distribution, but micro-probe “B” reveals more precise and more stable compared to micro-probe “A”. The characterization of the micro-probes is performed by comparing their measurement data to those obtained from the reference laser interferometer fixed in the same setup with the micro-probe and tracking the movable reflector bolted to the piezoelectric stage. On the opposite side of the stage, the micro-probe tracks the movement of a second reflector. Residual errors of the optical micro-probes, which correspond to the difference between the information given by the laser interferometer and the microprobe, are investigated with different metallic reflectors commonly

146

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

(a)

Stabilities of both optical micro-probles "A"

"B"

Residual errors vs. the displacement measured by the laser interferometer ("A") 0°

0.1

Residual errors (nm)

Micro-probes data (nm)

0.15

0.05 0 -0.05 -0.1 -0.15

40

0

0

5

10

15

20

25

2°

4°

6°

100 75 50 25 0 -25 -50 -75 -100 80

30

120

160

200

240

280

320

360

Laser interferometer data (nm)

Time (min) Fig. 7. Evaluation of the stabilities of the micro-probes (“A” and “B”) over a time range of 30 min.

(b)

Residual errors vs. the displacement measured by the laser interferometer ("B") 0°

Residual errors vs. the displacement measured by the laser interferometer ("A")

(a)

Aluminum

Ceramic

Gold

Copper

Steel

Silicon

Residual errors (nm)

used in dimensional metrology such as: aluminum, steel, silicon, copper, gold and ceramic materials. For a fixed working distance (WD) of 600 ␮m, calibration of the micro-probes is achieved by changing the gap over a measuring range of 360 nm. Micro-probe “A” exhibits residual errors lower than 15 nm except for test objects made of silicon or copper (Fig. 8(a)). However, for micro-probe “B”, the linear residuals are close to 2 nm except for test objects made of copper or gold deposits (Fig. 8(b)). The acceptance angle of the two micro-probes is investigated using a rotary stage having a range of ±6◦ . Since the results of a [0, +6]◦ tilting are symmetric to [−6, 0]◦ tilting, only those between

2°

4°

6°

90 60 30 0 -30 -60 -90 0

40

80

120

160

200

240

280

320

360

Laser interferometer data (nm) Fig. 9. (a) Evolution of the linear residuals for the micro-probe (“A”) when changing the gap over a travel range of 360 nm and the inclination of the target (0◦ , 2◦ , 4◦ and 6◦ ), (b) evolution of the linear residuals for the micro-probe (“B”) when changing the gap over a travel range of 360 nm and the inclination of the target (0◦ , 2◦ , 4◦ and 6◦ ).

Residual errors (nm)

35 25 15 5 -5 -15 -25 -35 40

0

80

120

160

200

240

280

320

360

Laser interferometer data (nm)

Residual errors vs. the displacement measured by the laser interferometer ("B")

(b)

Aluminum

Ceramic

Gold

Copper

Steel

Silicon

Residual errors (nm)

5 3 1 -1 -3 -5 0

40

80

120

160

200

240

280

320

360

0 and 6◦ are illustrated. According to the results in Fig. 9(a) and (b), the acceptance angle should be less than 2◦ , otherwise the residual errors become too high. Additionally, if the tilt angle exceeds 6◦ , the output signal is cut-off because the reflected laser beam does not return back to the beam splitter. Comparison of these results reveals that micro-probe “B” exhibits smaller residual errors than microprobe “A”, this error is found to be less than 2 nm when the tilt angle is kept less than 2◦ . In general, the discrepancy in the performances of the two micro-probes is ascribed to their different beam splitting ratios as dictated by the design of those micro-probes. Hence, values of the splitting ratio present a best fit with a specific range of reflectances. That is why micro-probe “B” performs better when measuring highly reflective test objects. To inspect the effect of roughness, three aluminum test objects have been selected, with the a 10-point mean roughness parameter Rz (ISO4287:1997) of: R1 = Rz = 0.02 ␮m (aluminum deposite) ( = 0.004 ␮m), R2 = Rz = 2.5 ␮m ( = 0.06 ␮m) and R3 = Rz = 4.5 ␮m ( = 0.15 ␮m), measured using the LNE ultra-high precision profilometer that is traceable to the SI meter definition [21]. The recorded residual errors were found to be proportional to the roughness of the aluminum surface as shown in Fig. 10(a) and (b). Nevertheless, micro-probe “B” exhibits smaller residual errors compared to micro-probe “A”.

Laser interferometer data (nm)

5.2. Second optical setup for micro-probe “B” Fig. 8. (a) Evolution of the linear residuals for the micro-probe (“A”) when changing the gap over a travel range of 360 nm and the kind of materials (aluminum, ceramic, gold, copper, steel and silicon), (b) evolution of the linear residuals for the microprobe (“B”) when changing the gap over a travel range of 360 nm and the kind of materials (aluminum, ceramic, gold, copper, steel and silicon).

For the second series of measurements, the laser diode has been replaced by a laser source tunable in the C–L bands. The tests are achieved only with micro-probe “B” since it is more accurate and

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

(a)

Residual errors vs. the displacement measured by the laser interferometer ("A")

(a)

R1

R2

Data

75 50 25

U_avr

0 -25 -50 -75

2 0 -2

-100

-4 0

40

80

120

160

200

240

280

320

360

0

50

Laser interferometer data (nm)

Residual errors vs. the displacement measured by the laser interferometer ("B") R1

R2

100

150

200

250

300

350

400

Laser interferometer data (nm) (b)

Impact of the working distance (WD) on the behaviour of the micro-probe ("B")

R3

WD = 600 µm

35

WD = 300 µm

U_ind

U_avr

4

25

Residual errors (µm)

Residual errors (nm)

U_ind

4

Residual errors (µm)

Residual errors (nm)

Residual errors vs. the displacement measured by the laser interferometer ("B")

R3

100

(b)

147

15 5 -5 -15 -25

2 0 -2

-35 0

40

80

120

160

200

240

280

320

360

Laser interferometer data (nm)

-4 0

60

120

180

240

300

360

Laser interferometer data (nm) Fig. 10. (a) Evolution of the linear residuals for the micro-probe (“A”) when changing the gap over a travel range of 360 nm and the roughness of the target (R1 = Rz = 0.02 ␮m (aluminum deposite)) ( = 0.004 ␮m), R2 = Rz = 2.5 ␮m ( = 0.06 ␮m) and R3 = Rz = 4.5 ␮m ( = 0.15 ␮m), (b) evolution of the linear residuals for the micro-probe (“B”) when changing the gap over a travel range of 360 nm and the roughness of the target as for the micro-probe (“A”).

more stable than micro-probe “A” when using highly reflective test objects. Characterization is performed step-by-step over a measuring range of 360 nm. To perform these tests, both the frequency and the intensity of the tunable laser source are fixed and only the wavelength is varied automatically between 1500 and 1600 nm, in steps of 5 nm. The calculated FFT provides the optical path difference; the residual errors instead are deduced by comparing these results to those provided by the reference laser interferometer. The obtained results for the residual errors lie in the micrometer range within ±4 ␮m and they exhibit a repeatable behavior as presented in Fig. 11(a). Varying the WD from 600 to 300 ␮m, the micro-probe exhibits a quite similar behavior as presented in Fig. 11(b). Based on these results, we conclude that the residual errors obtained with the technique of wavelength sweep (second setup) are much higher than those obtained from the fixed wavelength technique (first setup). However, it is worth to mention that the second setup offers the possibility of covering a much larger measurement range of several hundreds of micrometers. Numerous stability tests have been performed in order to interpret the reason underlying the high residual errors of ±4 ␮m. The temporal redundancy has been investigated for N points (N = 100pts, N = 250pts, N = 500pts, N = 750pts, N = 1000pts), but it does not impact the experimental results as presented in Fig. 12(a) and (b). Changing the intensity of the tunable laser from 1 to 4 W does not affect the behavior of the micro-probe too, as shown in Fig. 13(a) and (b). The last tests have been performed by changing the frequency of the tunable laser source from 1 to 4 THz as illustrated in Fig. 14(a). At the frequency of 3 THz and WD of 600 ␮m, the residual errors become lower than 1 ␮m. However, when the WD

Fig. 11. (a) Evolution of the linear residuals when changing the gap between the micro-probe (“B”) and the target: repeatability test (U ind: individual uncertainty, U avr: average uncertainty), (b) evaluation of the working distance on the linear residuals.

is reduced to 300 ␮m, the behavior of the micro-probe is slightly improved independently of the applied frequency as shown in Fig. 14(b). 6. Analysis and discussion The calibration of the developed micro-probes “A” and “B” is performed by comparing their data to those given by the RENISHAW laser interferometers using a specific test bench. The main advantage of the proposed approach is the use of an experimental setup with a nanometric level of accuracy that provides an ultra-precise qualification for the micro-probes. Therefore, we do not need any external standard to perform the calibration of the micro-probes. Another technique for calibrating the micro-probes consists on the use of an external standard with different U-grooves or stepped heights, calibrated on an ultra-high precision machine. The external standard is measured using the micro-probe and the recorded values are compared to the reference values. Increasing the number of steps degrades the uncertainty, but it can be sufficient when targeting calibration accuracy in the range of sub-micrometers. When targeting accuracy in the range of nanometers, this approach cannot be applied and we should compare the micro-probes output signals directly to the data given by the laser interferometer. Additionally, the environmental conditions should be controlled to compensate the influence of thermal expansion. The results obtained from the first optical setup shows that both micro-probes “A” and “B” exhibit accuracy in the nanometer range. But, the values of the residual errors are slightly affected by the material type because their reflectances are not identical. Similarly,

148

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

Stability of the micro-probe when changing the number of point (WD=600 µm)

(a)

100 pts

250 pts

500 pts

750 pts

4 Watt

1000 pts

6 4 2 0 -2 -4 -6

2 Watt

1 Watt

2

0

-2

-4 0

20

40

60

80

100

120

0

Time (min)

250 pts

500 pts

20

40

60

80

100

120

Time (min)

Stability of the micro-probe when changing the number of point (WD=300 µm)

(b)

750 pts

(b)

Stability of the micro-probe ("B") when changing the power (WD=300 µm)

1000 pts

4 Watt

4

3 Watt

2 Watt

1 Watt

6

Residual errors (µm)

Residual errors (µm)

3 Watt

4

Residual errors (µm)

Residual errors (µm)

Stability of the micro-probe ("B") when changing the power (WD=600 µm)

(a)

2

0 -2 -4 0

20

40

60

80

100

120

Time (min)

4 2 0 -2 -4 -6 0

20

40

60

80

100

120

Time (min)

Fig. 12. (a) Investigation of the micro-probe stability for a working distance equal to 600 ␮m. Each plotted point in this graph corresponds to the average of N points (N = 100 pts, 250 pts, 500 pts, 750 pts and 1000 pts), (b) investigation of the microprobe stability for a working distance equal to 300 ␮m. Each plotted point in this graph corresponds to the average of N points (N = 250 pts, 500 pts, 750 pts and 1000 pts).

Fig. 13. (a) Investigation of the micro-probe stability when changing the intensity of the output signal: 1 W, 2 W, 3 W and 4 W (working distance equal to 600 ␮m), (a) investigation of the micro-probe stability when changing the intensity of the output signal: 1 W, 2 W, 3 W and 4 W (working distance equal to 300 ␮m).

both the roughness and the tilting of the target affect the behavior of the tested micro-probes “A” and “B”. Hence when the target is too rough, the intensity of the reflected laser beam becomes too low, which influences its behavior. To acquire an acceptable behavior from the micro-probes, the target roughness should be less than 1 ␮m. To respect the acceptance angle, the tilt angle between the micro-probe and the target should be less than 2◦ . Otherwise, the intensity of the reflected laser beam becomes also too low. When the tilt angle exceeds 6◦ , the output signal is cut-off since the reflected laser beam cannot reflect back onto the beam splitter. Testing micro-probes “A” and “B” under the same conditions reveals that micro-probe “B” (corresponding to a beam splitter thickness of 3.62 ␮m and reflectance–transmittance of 25–75%) is more accurate than micro-probe “A” (corresponding to a beam splitter thickness of 3.58 ␮m and reflectance–transmittance of 75–25%). This means that the thickness of the beam splitter – which has an impact on the reflectance–transmittance ratio – should be optimized to ensure low reflectance ratio of 25% and high transmittance ratio of 75%. Based on the presented results in section (IV-A), micro-probe “B” seems to be more accurate and appropriate for dimensional metrology applications with a nanometer target uncertainty such as measurements of straightness, roundness and cylindricity (form errors [20]) as well as the measurement of profile and roughness [21] inside small bore with a diameter inferior to 1 mm. The contactless calibration of the form errors and the roughness of small bore represent a challenging request in NMIs (calibration of standard artifacts) and in industrial applications, related to:

automotive (fuel injection nozzles), horology (small gears and parts), spatial, and medical (needle-free injection system), etc. In general, improvement of the measurement uncertainty leads to enhancement of the manufacturing process, which in turn, allows obtaining high quality surfaces. As an example, mastering the nozzle bore shape leads to better control on the fuel injection dose and to reduce the waste of energy. According to the presented results, the measuring range of both micro-probes is 360 nm which is considered to be very low for dimensional metrology applications. For such kind of applications, the form errors of the small bore are usually less than 1 ␮m. In addition, it is essential to take into account the eccentricity between the artifact-axis and the measuring axis that can be adjusted to less than 1 ␮m, together with the motion errors of the mechanical guiding systems translating the metrology frame where the micro-probe will be integrated (measurement of profile and roughness). For roundness measurement, only the artifact undergoes the motion errors provoked by the high precision spindle equipping the measuring machine and characterized by the fixed micro-probe [20]. The motion errors of the guiding systems can be estimated to few micrometers. In conclusion, the measuring range of the micro-probe should be raised at least to 10 ␮m. This way, the characterization tests are performed again with the second optical setup starting with a fixed frequency and intensity of the tunable laser source. For each stepped motion of the reflector, only the wavelength is changed step-by-step automatically in the range of 1500–1600 nm. The obtained results show that the accuracy of the micro-probe degrades to the micrometer range. The

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

(a)

(a)

Stability of the micro-probe when changing the frequency bande-width (WD=600 µm) 2 THz

3 THz

2

4 THz

Signal (V)

1 THz

2 1

1 λ 1500 nm λ 1550 nm λ 1600 nm Intersection

0 -1

0

-2 365

-1

367

369

371

373

375

Travel range (µm)

(b)

-2

2

-3 0

20

40

60

Time (min)

80

100

120

Stability of the micro-probe when changing the frequency bande-width (WD=300 µm)

(b)

1 THz

2 THz

3 THz

4 THz

Out-put signal (V)

Residual errors (µm)

3

3

1

λ (1500) λ (1500)

0

λ (1600) λ (1550)

-1

λ (1600) λ (1600)

fit

fit

fit

-2 0

500

2 1 0 -1 -2 -3 0

20

40

60

80

100

120

Time (min) Fig. 14. (a) Investigation of the micro-probe stability when changing the wavelength around 1500 nm; the working distance is equal to 600 ␮m; and four tests are performed with four fixed frequency are equal 1 THz, 2 THz, 3 THz, 4 THz, (b) investigation of the micro-probe stability when changing the wavelength around 1500 nm; the working distance is equal to 300 ␮m; and four tests are performed with four fixed frequency are equal 1 THz, 2 THz, 3 THz, 4 THz.

aforementioned accuracy may be attributed to the tunable laser source, in particular, the mechanical rotation of the Bulk grating, the minute changes in the position of the optical elements caused by mechanical drifts over time that can rapidly degrade the power, the spectral purity, the electronic noise and the FFT analysis. The tests performed by varying the wavelength within 1500 and 1600 nm reveals that the accuracy of the micro-probe is still in the micrometer level, but they also demonstrate that the microprobe can still be used at different wavelengths lying in the range of 1500–1600 nm. A numerical simulation using Matlab validates the fact that the employment of three laser diodes (at the wavelength of 1500, 1550 and 1600 nm) improves the travel range to 372 ␮m as shown in (Fig. 15(a)). This travel range corresponds to the case of perfect signals and is too high compared to the travel range of 360 nm obtained when using only one laser diode with a wavelength of 1550 nm (Fig. 15(b)). A second simulation has been performed considering the three wavelengths of 1500, 1550 and 1600 nm combined with the noise signal identified experimentally through the first optical setup. After converting the analog signal (raw and fitted signals) to distance signal (Fig. 15(b)), the results confirm that the travel range of the micro-probe can reach 372 ␮m. For any three selected analog values, presenting a similar abscissa, it is possible to find the distance for the entire measuring range of 372 ␮m. An example with simulation measurement range overcoming 3200 nm is shown in the Fig. 15(c).

1000 1500 2000 2500 3000 3500

Distance (nm)

(c)

Converted out-put signal to distance (nm)

Residual errors (µm)

149

200 100

λ 1500 λ 1500

0

λ 1550 λ 1550

-100

λ 1600 λ 1600

-200 0

Fit

Fit

Fit

Abscissa Values

500 1000 1500 2000 2500 3000 3500

Distance (nm) Fig. 15. (a) Simulation: identification of the first intersection point when using three analogical sinusoidal wavelength of 1500, 1550 and 1600 nm. The abscissa of the first intersection point is equal to 372 ␮m. The magnitude of the signals does not influence the abscissa of the intersection point, (b) simulation: reconstruction of three analogical sinusoidal signals basing on the experimental signals recorded when using the laser diode of 1550 nm (: wavelength 1500 nm, 1550 nm and 1600 nm, the symbol “Fit” indicate the fit of each reconstructed signal to eliminate the noise), (c) conversion of the simulated analogical signals to distance signals. The abscissa represents the distance given by the micro-probe which can vary until a maximal travel range of 372 ␮m. As example: for a three magnitude values of 126.5 (signal of 1500 nm), −1.11 (signal of 1550 nm) and 86.58 (signal of 1600 nm) the deduced distance is equal to 3336 nm.

7. Conclusion The characterization of two optical silicon micro-probes “A” and “B” with two thicknesses of the beam splitter is achieved, thanks to the developed ultra-high precision test bench, under a controlled environment. Two distinguished optical setups are proposed and evaluated for the micro-probes. The first setup is based on the use of a laser diode at the wavelength of 1550.3 nm and the second is based on the use of a tunable laser source. Applying the first optical setup reveals that the micro-probes exhibit a nanometer level of accuracy for a travel range of 360 nm. Results prove that the micro-probes are sensitive to the tilting and the roughness of the target especially when they exceed 2◦ and 1 ␮m respectively; however, they are less sensitive to the material type. When applying the second optical setup, the tests are performed under a variable wavelength between 1500 and 1600 nm. As a consequence, the residual errors increase to the micrometer range, but a large travel range can be reached. To improve both the linear residuals and the travel range, a simulation of the first optical setup is performed considering three separate laser wavelengths

150

H. Nouira et al. / Sensors and Actuators A 223 (2015) 141–150

of 1500, 1550 and 1600 nm. The obtained result demonstrates that the travel range can be improved to 372 ␮m. References [1] A. Weckenmann, T. Estler, G. Peggs, D. McMurtry, CIRP Ann. Manuf. Technol. 53 (2004) 657–684. [2] H. Haitjema, W.O. Pril, P.H.J. Schellekens, CIRP Ann. Manuf. Technol. 1 (2001) 365–368. [3] G.N. Peggs, A.J. Lewis, S. Oldfield, CIRP Ann. Manuf. Technol. 1 (1999) 417–420. [4] F. Meli, M. Fracheboud, S. Bottinelli, M. Bieri, R. Thalmann, J.M. Breguet, R. Clavel, EUSPEN Int. Topical Conf, 2003. [5] V. Nesterov, P. Pornnoppadol, U. Brand, R. Wilke, M. Schmidt, S. Buettgenbach, SPIE Conf. Smart Sens., Actuators and MEMS 5116 (2003) 844. [6] M. Petz, R. Tutsch, R. Christoph, M. Andraes, B. Hopp, Measurement 45 (2012) 2288–2298. [7] R. Onodera, Y. Ishii, Opt. Commun. 167 (1999) 47–51. [8] D.M. Silva, E.A. Barbosa, N.U. Wetter, Rev. Sci. Instrum. 83 (2012), 103103, 7 pp. [9] P. Giacomo, Metrologia 20 (1984) 25–30. [10] M. Malak, F. Marty, H. Nouira, G-P. Vailleau, T. Bourouina, Appl. Phys. Lett. 102 (2013), 141102, 5 pp. [11] M. Malak, F. Marty, H. Nouira, J. Salgado, T. Bourouina, Int Conf. on Micro Electro Mechanical Systems, 628, 2012. [12] M. Malak, F. Marty, H. Nouira, J. Salgado, T. Bourouina, Int Conf. on Optical MEMS and Nanophotonics (OMN), 109, 2012. [13] http://www.vlsistandards.com/products/dimensional/lattice.asp?SID=78. [14] H. Nouira, R. Bergmans, A. Kueng, H. Piree, R. Henselmans, H. Spaan, Int. J. Metrol. Qual. Eng. 5 (2014), 204, 13 pp. [15] S. Yokoyama, T. Araki, N. Suzki, Rev. Sci. Instrum. 66 (1995) 2788–2795. [16] T. Eom, H. Choi, S. Lee, Rev. Sci. Instrum. 73 (2002) 221–224. [17] J. Stone, J. S.D. Phillips, Res. Natl. Inst. Stand. Technol. 101 (1996) 671–674. [18] K.P. Birch, M.J. Downs, Metrologia 31 (1994) 315–316. [19] R.K. Leach, Fundamental Principles of Engineering Nanometrology, William Andrew Publishing, UK, 2009. [20] A. Vissiere, H. Nouira, M. Damak, O. Gibaru, J.M. David, Meas. Sci. Technol. 23 (2012), 094014, 9 pp. [21] H. Nouira, J-A. Salgado, N. El-Hayek, S. Ducourtieux, A. Delvallée, N. Anwer, Meas. Sci. Technol. 25 (2014), 044016, 12 pp.

Biographies

Dr. Hichem Nouira received the Ph.D. degree in 2008 at Université de Franche Comté. Since 2008, he took the position of research scientist at the LNE and project leader of the French research programme focused on the manufacturing of a new machine for cylindricity measurement at the nanometer level of accuracy. Currently he involved in four EMRP projects (“FORM” (ind10), “THERM” (ind13), “MICRO-PARTS” (ind59) and “TIM” (ind62)). He gained new competence in the dimensional metrology, development of machines with ultra-high precision for metrology applications, optical measurement and data processing.

Dr. Jean-Pierre Wallerand received the Ph.D. degree from the Conservatoire National de Arts et Métiers (CNAM), Paris, France, in 1997, for works on high resolution spectroscopy of molecular iodine particularly resonant Raman spectroscopy. Since 1998, he has been a Research Scientist at CNAM, working in the dimensional metrology department. He is currently in charge of the research activity of the dimensional department of the joint laboratory LNE-CNAM (LCM). He is involved in development of long distance measurement systems since 2007 and in the implementation of interferometric microprobes, using frequency scanning interferometric or multiwavelength interferometry since 2012.

Dr. Maurine Malak received the B.Sc. degree in electronics and communication engineering and the M.Sc. degree in integrated optics from Ain Shams University, Cairo, Egypt, in 2004 and 2008, respectively, and the Ph.D. degree in optical MEMS from Paris-Est University, France, in 2011. From 2004 to 2008, she was a Research Assistant with the optical communication and Laser Lab, Ain Shams University. From 2006 to 2008, she was a part time teaching Assistant with the French University, Egypt. From 2008 to 2011, she was a part-time Teaching Assistant in ESIEE Paris, France. In 2012, she was a Postdoctoral Fellow in ESIEE Paris, France. From 2012 to 2014, she was a Postdoctoral Fellow with the Ecole Polytechnique Federale de Lausanne, Switzerland. During her career, she worked on the design and characterization of integrated optical devices for CWDM systems and for near-infrared spectrometry for space applications. She also worked on the design and implementation of photonic MEMS devices intended to spectrometry and profilometry applications. She developed a novel lab-on-chip complex refractometer. She is the author of a book chapter about MEMS Deep 1-D Photonic crystals, and more than 35 journal papers and conference proceedings. Her current research interests include the design and implementation of optical devices for applications targeting lab-on-chip systems. She is also interested in optofluidics, optomechanics, optical mems, integrated optics, and geometrical optics. She received the Best Thesis Award from the Descartes Development Agency in 2012, student mobility Grant from Paris-Est University in 2010, and the French Government Grant in 2008–2009. She is affiliated to the AERES national French Community, dedicated to the evaluation of the French laboratories. She is also Member of the SPIE scientific society. Dr. Anne-Franc¸oise Obaton received the Ph.D. in Physics from University of La Rochelle, France, in 1998 for her research work on “New Yb3+ -Er3+ codoped phosphate glasses for eye-safe laser applications”. In 1999, she had a post-doctoral position in the Institute of Industrial Science, Laboratory for Integrated Micro-Mechatronic Systems in Tokyo University, Japan, working on AFM. In 2008, she received her habilitation to supervise research from University Pierre et Marie Curie, Paris, France. Since 2000 she has been involved in metrology at the Laboratoire national de métrologie et d’essais (LNE) in Paris, France. Her research interest involves optical sensors and additive manufacturing. Dr. José Salgado, has 14 years of experience in dimensional metrology, since 2013 he is the technical and R&D head of the dimensional metrology department. He is the responsible for interferometric measurements, linescales, stage micrometer measurements and surface texture. He has successfully developed various precision measuring instruments such as a stage micrometer calibration instrument and 3D roughness measuring instrument. Since 2013 he is the French contact person at EURAMET TCL committee.

Prof. Tarik Bourouina holds a Master of Science (Physics), a Master of Engineering (Electronics) and the Ph.D. degree, obtained in 1991 at Université Paris XII. His career started in 1988 and was entirely devoted to the field of MEMS sensors and analytical opto-fluidic micro-instruments. Dr. Bourouina took several academic positions in France and in Japan, at the Université Paris-Sud Orsay, at the French National Center for Scientific Research (CNRS) and at The University of Tokyo. Dr. Bourouina is now full Professor at ESIEE Paris, Université Paris-Est.