Development of a Low Noise External Cavity Diode Laser in the Littrow Configuration: Progress Report #1 Chloe Ling LIGO SURF 2013 Mentors: Rana Adhikari and Tara Chalermsongsak July 8, 2013 I have spent the first 3 weeks of the summer designing my external cavity diode laser (ECDL) and reading literature in order to understand the theory behind my project. This involves determining the acceptable noise level, figuring out which parts will meet this requirement, and sketching a reasonable experimental setup.

1

Background

The Laser Interferometer Gravitational-Wave Observatory (LIGO) relies on stable laser light for interferometric measurements. However, most stable lasers are very expensive. External cavity diode lasers (ECDLs) have been shown to be a much cheaper alternative. Laser diodes work by running current through the p-n junction of a semiconductor diode. Single mode laser diodes have a built in mechanism of wavelength selective loss, where the selected wavelength of the laser diode has the lowest loss creating coherent light [1]. However, laser diodes have high noise levels and large linewidths, making them poor candidates for high precision interferometric measurement. ECDLs work by reducing the noise level of a laser diode through optical feedback in an external cavity. The purpose of this project is to design, build, and test a 1064 nm ECDL for use by LIGO. Two configurations are commonly used for ECDLs: the Littrow configuration and the Littman-Metcalf configuration. The Littrow configuration uses a movable diffraction grating to reflect first order beams back into an external cavity. In contrast, the Littman-Metcalf configuration has a fixed diffraction grating with a movable mirror to tune for the correct output wavelength. We chose to use the Littrow configuration instead of the Littman-Metcalf configuration because the Littrow configuration has been shown to have higher ouput power, since the output beam is directly reflected light [2]. The Littman-Metcalf configuration is also more difficult to tune correctly since the diffraction grating is immovable. Finally, there are many more exemplars of the Littrow configuration in literature. In our setup, based on the Littrow configuration, the external cavity is formed between the back of the lasing material and a diffraction grating. Light from the laser diode is sent through a collimating lens, which makes the beam travel in a single direction. This light reaches a diffraction grating. Only first order beams at a wavelength determined by the angle of the grating are reflected back into the cavity. This process of optical feedback continues, which reduces the noise level since only one wavelength can continue to travel in the cavity. Zero order beams are reflected off the diffraction grating and help contitute the beam exiting the ECDL [3]. 1

We can improve the frequency noise suppression further by locking the ECDL to a reference cavity. We can actuate on the frequency noise by tuning the reference cavity with a piezoelectric actuator and using current feedback. The ECDL will be temperature controlled with a thermoelectric cooler since the lasing material is temperature sensitive.

2

Design

2.1

Noise Requirements

In order to create a useful ECDL for LIGO, it must be capable of meeting the noise requirements of the experiment it is being used for. Thus, we examined the noise levels of the possible experiments we could perform with our ECDL to choose which experiment we will use to test it. • NPRO: [4] • Crackle experiment: LIGO has mirrors hanging from suspensions, which experience crackling noise. This is noise due to the mechanical apparatus being driven at low frequencies resulting in high frequency excitation, or ’crackling’. This work seeks to study how crackling noise can affect a low noise mechanical apparatus [5]. The experimental setup for the crackling experiment involves a Michelson interferometer with blade springs. The blade springs are driven in unison at low frequency, and any noise that√arises is from ’crackling’ and can be observed. We estimated the noise level of 300Hz/ Hz based on the shot noise limit of the setup [6]. • CTN and Cryo experiments: These experiments both aim to measure length noise in dual reference cavities. The CTN experiment is conducted at room temperature, and the Cryo experiment is conducted at cryogenic temperatures. In the experiment, 2 lasers are each frequency locked onto one of 2 reference cavities. The two transmitted beams are recombined, creating a beat frequency. The noise on the beat frequency can be used to calculate the length noise√of the 2 cavities. The frequency noise requirement 0.1 for CTN is approximately √ Hz/ Hz and the frequency noise requirement for Cryo is f √ √ Hz/ Hz. We take the servo gain into account, where we have a gain approximately 0.03 f

of

1 f3

up to 10 kHz, and

1 f

up to 1 MHz.

In order to perform the CTN and Cryo experiments, the ECDL needs to be able to lock to a high finesse reference cavity. Tara determined that √ for this to be possible, the ECDL needs to have a frequency noise of less than 400Hz/ Hz [7]. • HeNe laser: In this setup, a Fabry-Perot cavity is formed with one entirely reflective mirror and one partially reflective mirror on opposite sides of a cavity filled with helium and neon gas. Note that the HeNe data presented here is after frequency stabilization, and the actual data has significantly higher noise levels [8]. • Birmingham group: The Birmingham group built a 1064 nm ECDL, and we examine how their noise levels compare to the requirements we may use to test our ECDL [9]. They were able to achieve rather low noise levels, and we hope to be able to recreate this with our device.

2

Based on the possible experiments we could perform, we have settled on using the Crackle experiment since it has the highest noise requirement of the experiments we examined and will therefore be the easiest standard to meet. Plotted below, we have the noise levels of all of the mentioned experiments. Also included are the noise levels due to temperature and current discussed in Section 2.2.

2.2

Estimated Noise of Bare Diode Laser

First, we estimated the intrinsic noise of the diode laser (and thus its predicted linewidth). We considered two main sources of noise in order to make this prediction: 1. Noise from current: Some of the laser diode noise comes from a varying current input. We will likely be using a low noise current driver based on the Libbrecht and Hall design [10], or a modified commercial driver that will have low enough noise for our purposes. The √ Libbrecht and Hall design has a PSD noise of 200 pA/ Hz. The oscillation frequency of a GaAs laser changes due to current fluctuations √ at 3 GHz/mA [11]. We conclude that the noise from current fluctuations is 6 ∗ 102 Hz/ Hz. 2. Noise from temperature: We plan to use the TEC built into the laser diode mount from Thorlabs √ HLD001. We will have some intrinsic noise from the laser, of about 2 ∗ 10−4 /f K/ Hz. The oscillation frequency of a GaAs laser changes due to temperature fluctuations at about 20 GHz/K [11]. We conclude that the noise from temperature √ fluctuations is 4 ∗ 106 /f Hz/ Hz. q

The noise in power spectral density (PSD) adds in quadrature; that is, S = S12 + S22 , so √ √ 3.6∗105 f 2 +1.6∗1013 the total noise will be S(f ) = Hz/ Hz. This gives us an estimate of the f 3

noise in a PSD spectrum.

2.3

Estimated Noise of Diode Laser with External Cavity

As discussed before, the external cavity reduces noise by optical feedback. In order to estimate the factor by which the noise is reduced, we calculate parameter X [11, p.963]. l1 is the length of the laser cavity, l2 is the length of the external cavity, S is the scattering matrix for the laser, and Γ1 and Γ2 are the complex amplitude reflection coefficients. From Saito, X=

τ1 τ1 Ref f 1/2 |S12 S21 Γ2 /S11 | = ( ) τ2 τ2 R2

where τ1 = 2lc11 , τ2 = 2lc02 , Ref f is the effective reflectivity of the external grating, and R2 is the laser diode reflectivity. Simplifying, we find l2 Ref f 1/2 X=n ( ) l1 R2 These dimensions are all determined by the choice of laser, grating, and cavity length (done explicitly in Section 2.5). Therefore, we chose what components to purchase by modeling the noise they would generate. Most of the laser diodes on the market are GaAs, with n ≈ 3.5, so we used this approximate value in our calculation. l1 was found for each laser diode based on the dimensions given by the manufacturers. We began with l2 ≈ 10cm since the Birmingham group was successful at reducing noise levels with this parameter, and modeled different variations from this value [9]. Ref f was determined by the efficiency of the grating output, which we estimated using the efficiency plots given by Thorlabs. R2 for GaAs was found to be about 0.85 [12]. From Saito [11, p.965], we have an equation which predicts how the noise transforms in power spectral density: S E (f ) =

(1 + X sin(f τ2 )/(f τ2

))2

S(f ) + X 2 (1 − cos(f τ2 ))2 /(f τ2 )2

This reduction of noise is independent of frequency for small frequencies, where f τ2 ≤ 0.3 [11, p.965], where f is the oscillation frequency in the external cavity. When f τ2 becomes close to 1 (we approach the free spectral range), we see an increase in noise due to the jump from one axial mode to another in the external cavity. This noise could present a problem in being able to lock the ECDL to a high finesse reference cavity, since we are limited by the current noise of the laser diode.

2.4

Estimated Noise of ECDL with servo

Next, we are interested if the addition of a servo will be sufficient to reduce our noise to the levels needed to perform our experiment. We can estimate a transfer function that goes as f13 up to 10 kHz, and as f1 up to 1 MHz, similar to what is used for Cryo. This has a gain of G=

1014 f3

+

106 f .

Thus, our final noise level can be reduced by the servo by a factor of Sf inal =

4

SECDL 1+G

2.5

Choosing Components

Using all of this information, we can model how the specifications of different components of the ECDL (diodes, gratings, cavity length) will affect the output noise level in PSD. First, we examine how different diodes affect the final noise level. The diodes that are on the market now are most commonly GaAs for 1064 nm, so the only parameter that affects noise is the size of the diode chip. Specifically, l1 from the parameter X calculation.

Based on this graph, we plan to order the Thorlabs and QPhotonics diodes to measure their PSD and compare them. The diode from Lumics has suspiciously low noise, and its package does not fit into the same mount as the Thorlabs and QPhotonics diodes, so we will not be ordering this diode. The companies are not able to guarantee an exact wavelength of 1064nm to their diodes, but the diodes are tunable with temperature, at 0.3nm/◦ C. Next, we compare different grating choices. Grating choices affect Ref f in the parameter X calculation, but the groove spacing does not appear explicitly in parameter X. There is also not a specific correlation between effective reflectivity and groove spacing. We examine different gratings with 2 different blaze wavelength to see how their efficiencies affect the noise levels. In this case, efficiency refers to how much power is reflected off the grating relative to the incident power.

5

Based on this, it would make the most sense to choose the grating with the highest efficiency. However, the groove spacing affects cavity dispersion (the sensitivity of the output wavelength due to different grating angles) [1]. The best cavity dispersion occurs when the the groove spacing is close to λ2 , so we will go with the Thorlabs 1200/mm 1 µm blaze as our grating choice. We also compare the noise levels for different cavity lengths. As we see, increasing the cavity length decreases the noise level, but also decreases the free spectral range. This means the noise rises again at a lower frequency, and could affect our ability to lock the ECDL to a reference cavity.

6

Based on this, we will choose a cavity length between 6 and 10 cm. We also need to order a diode mount, a thermoelectric cooler (TEC) and TEC controller to maintain a constant temperature, and a low noise current driver. We are trying to find a current driver in the lab that we can modify to meet our needs. The TEC and TEC controller are both parts we need to order. Thus, we conclude the following parts list. • Thorlabs M9-A64-0200 laser diode: 200 mW, GaAs, 9 mm package, $772.44 • QPhotonics QLD-1060-100s laser diode: 100 mW, InGaAs, 9 mm package, $240.00 • Thorlabs GR25-1210 diffraction grating: 1200 grooves/mm, 1 µm blaze, 25x25x6 mm, $102.00 • TEC: We will be buying a TEC controller from Thorlabs as well as a TEC that we will connect ourselves to the ECDL • Current controller: There is not currently a spare current driver in the LIGO labs. I am going to look for a commercial board and put this into a standard LIGO module.

2.6

Experimental Setup

The following experimental setup will be mounted inside a heavy, rigid metal box. Ideally, we will find a box with a lid that screws on tightly that will minimize mechanical and thermal noise from the surroundings. Attached to the laser diode will be the TEC element and this will connect to the TEC controller and current driver outside the box.

7

Our experimental setup is based heavily on Ricci et al. [13]. The angle of the grating is determined by the grating equation: mλ = d(sin θi + sin θr ) where θi and θr are the incident and reflected beams, measured counterclockwise from normal to the grating. In our case, our ECDL is for 1064 nm light, so λ = 1.064 ∗ 10−6 m. An ECDL in the Littrow configuration works by reflecting first order beams (m = 1) back into the external cavity, so we want θi = θr . Our grating choice is 1200 grooves/mm, so d = 1/1200 mm. From this, we conclude that θ = 39.7◦ . We check how the angle of the output beam is affected by a change in frequency. By differentiating the grating equation, dθ −mc = df 2df 2 cos θ 8

so for a tuning range of ∆f ≈ 100M Hz − 4GHz, ∆θ ≈ 1µrad. This is small enough that the change in angle at different frequencies will not affect the direction of the output beam significantly, so an additional mirror in the setup is unnecessary.

3

Future Plans

We have placed orders for all of the parts listed in Section 2.5. While I wait for them to arrive, I have several tasks to complete in the next couple of weeks: • Find a box that the ECDL will be built in. This box also needs to have a lid that can be tightly screwed on. The material needs to be very rigid and thick in order to avoid mechanical noise entering the system. First I will search the 40m lab, then perhaps online if I cannot find something that meets our needs. • Build the element which the grating will be glued to that contains the PZT. This involves building the element in Solidworks then sending the design plan to be machined. • Modify one of the current drivers in the Cryo lab so that it can be used to test the laser diode when the part arrives. Ideally, it will be very low noise. This is preferred to buying a current driver modeled after the Libbrecht and Hall design [10], since it would be quite expensive. • Figure out exactly how we will wire the TEC and current elements onto the laser diode so that I can begin constructing the ECDL as soon as we have all the parts. • Figure out how to measure the frequency noise once we have a working ECDL. There are multiple methods I need to look into here. The main idea is that I will beat the output beam from the ECDL with a very stable 1064 nm laser in the Cryo lab - by measuring this beat frequency I can figure out how much noise there is.

References [1] B. Mroziewicz. External cavity wavelength tunable semiconductor lasers - a review. OptoElectronics Review, 16(4):347–366, 2008. [2] R. E. Scholten C. J. Hawthorn, K. P. Weber. Littrow configuration tunable external cavity diode laser with fixed direction output beam. Review of Scientific Instruments, 72(12):3, 2001. [3] M. G. Boshier A. S. Arnold, J. S. Wilson. A simple extended-cavity diode laser. Review of Scientific Instruments, 69(3), 1998. [4] Rick Savage Rich Abbott, James Mason. Npro frequency stabilization. LIGO Technical Note, February 1997. [5] Rana Adhikari Eric Gustafson. Experiments to investigate creak noise in mechanical flexures and joints. LIGO Technical Note, December 2010. [6] Tara Chalermsongsak. Frequency noise requirement for laser used in crackle experiment. ELOG: PSL lab entry, April 2013. 9

[7] Tara Chalermsongsak. Laser noise requirement for locking to a refcav. ELOG: PSL lab entry, June 2013. [8] Akito Araya. Master’s thesis, The University of Tokyo, 1992. [9] University of Birmingham. Low-cost 1064nm light source for table-top interferometry based on a frequency stabilized external cavity diode laser, LVC meeting, Bethesda, Maryland, USA, March 2013. [10] J. L. Hall K. G. Libbrecht. A low-noise high-speed diode laser current controller. Review of Scientific Instruments, 64(8):2133–2135, August 1993. [11] Yoshihisa Yamamoto Shigeru Saito, Olle Nilsson. Oscillation center frequency tuning, quantum fm noise, and direct frequency modulation characteristics in external grating loaded semiconductor lasers. IEEE, QE-18(6):961–970, June 1982. [12] Robert E. Scholten Sebastian D. Saliba. Linewidths below 100 khz with external cavity diode lasers. Applied Optics, 48(36):6961–6966, December 2009. [13] T. Esslinger A. Hemmerich C. Zimmermann V. Vuletic W. Konig T. W. Hansch L. Ricci, M. Weidemuller. A compact grating-stabilized diode laser system for atomic physics. Optics Communications, 117:541–549, June 1995.

10