## Abstract

We explore optical-to-terahertz conversion efficiencies which can be achieved with femto- and picosecond optical pulses in electro-optic crystals with periodically inverted sign of second-order susceptibility. Optimal crystal lengths, pulse durations, pulse formats and focusing are regarded. We show that for sufficiently short (femtosecond) optical pulses, with a pulsewidth much shorter than the inverse terahertz frequency, conversion efficiency does not depend on pulse duration. We also show that by mixing two picosecond pulses (bandwidth-limited or chirped), one can achieve conversion efficiency, which is the same as in the case of femtosecond pulse with the same pulse energy. Additionally, when the group velocity dispersion of optical pulses is small, one can substantially exceed Manley‒Rowe conversion limit due to cascaded processes.

©2006 Optical Society of America

## 1. Introduction

Optical rectification (OR) of ultrashort laser pulses is an established way of generating broadband terahetz (THz) radiation. THz output in this case is produced in an electro-optic (EO) medium via difference-frequency mixing between Fourier components of the same optical pulse. First demonstrated with picosecond pulses in ZnTe, ZnSe, CdS and quartz [1] crystals, as well as in LiNbO_{3} [1,2], this technique was later extended to femtosecond laser pulses [3] which allowed generation of much broader bandwidths in terahertz region, [4] and even reaching mid-IR freqiencies of 40 to 50 THz [5,6].

Typically, to generate broadband THz transients via OR, thin (~1mm or less) electro-optic crystals are used, because of phase matching constraints as well as high absorption in conventional crystals (LiNbO_{3}, ZnTe) at THz frequencies. Optical-to-THz conversion efficiencies achieved so far by optical rectification methods are low [7], typically 10^{-6} - 10^{-9}, even with femtosecond pump pulse energies as high as 10 mJ [8]. However, recently conversion efficiency of 5×10^{-4} was reported. The authors used OR in LiNbO_{3} with tilted pulse front excitation and energy per optical pulse 500 μJ [9].

In order to enhance the optical-to-THz conversion efficiency, larger interaction lengths with collinear interaction of THz and optical waves is desirable. The idea of using quasi-phase-matched (QPM) materials with periodically-inverted sign of second-order susceptibility *χ*
^{(2)} for THz optical rectification was proposed by Lee and coauthors [10]. Multi-cycle narrow-band terahertz radiation was produced in periodically poled lithium niobate (PPLN) crystal. The authors used femtosecond pulses at 800 nm and cryogenically cooled (18K) PPLN crystal to reduce THz absorption, and achieved 10^{-5} conversion efficiency. OR in this case gives rise to a THz waveform which corresponds to the domain structure of the PPLN. Recently, efficient narrow-band terahertz radiation was demonstrated in another QPM material, orientation-patterned gallium arsenide (OP-GaAs) [11], which is extremely promising for THz generation because of intrinsically small THz absorption in GaAs.

THz wave generation in QPM crystals is illustrated in the time-domain animation of Fig. 1. Each inverted domain of a nonlinear crystal contributes a half-cycle of the THz pulse [10] and thus the THz wave packet (provided that THz-wave attenuation is low) has as many oscillation cycles as the number of quasi-phase-matched periods over the length of the crystal.

The purpose of this paper is to evaluate maximum optical-to-terahertz conversion efficiency which can be achieved by quasi-phase-matched down conversion process and to find optimal conditions in terms of pump wavelength, pump pulse format, crystal length, and focusing.

## 2. Plane wave analysis, femtosecond pulses (optical rectification).

Consider as an optical pump, bandwidth-limited ultrashort (e.g., femtosecond) laser pulses propagating along *z* in the form of infinite plane waves, with the gaussian time envelope of the electric field

where ω_{0} is the central frequency and τ is the pulsewidth.

Intensity envelope is thus *I*(*t*) ~ exp(-2*t*
^{2}/*τ*
^{2}) and the pulse duration at full width of half-maximum is *τ*
_{FWHM}=(2ln2)^{1/2}
*τ*= 1.18*τ*. Using Fourier transform pair in the form

and taking into account that *E*(*t*) is real, we can get (z=0) a transform of the electric field (1) in the form $\frac{1}{2}\left\{E\left(\omega \right)+{E}^{*}\left(-\omega \right)\right\}$, where *E*(ω) is a one-sided (ω > 0) Fourier component given by

Similar form of one-sided Fourier representation will be used for other fields below in the text. For arbitrary *z*,

where *k*(ω) is a module of the wave vector.

The one-dimensional equation (scalar form) for the Fourier component of the THz field at the angular frequency Ω,*E*(Ω,z), follows directly from Maxwell’s equations and is given, in the slowly varying envelope approximation and in the limit of no absorption, by [12]:

where the Fourier component of nonlinear polarization, *P*_{NL}
(Ω), can be expressed through the material nonlinear susceptibility *χ*
^{(2)} as

Here ε_{0} and μ_{0} are respectively the permittivity and permeability of free space, *c* is the speed of light in vacuum, *n*
_{1} is the THz refractive index. From Eqs. (4) and (7) it follows that

and (6) becomes

The *k*- vector mismatch Δ*k* is given by the following relation

Since Ω ≪ ω, we can replace *k*(*ω* + Ω) - *k*(*ω*) by (*∂k*/*∂ω*)_{opt} Ω. [1,13] and obtain Δ*k* in the form

where *n*
_{THz} is the phase refractive index for the THz wave, ${n}_{\mathit{\text{opt}}}^{\mathit{\text{gr}}}$
is the optical group velocity refractive index, and ʌ is the QPM orientation-reversal period. We assumed here that the optical group velocity dispersion $\frac{{\partial}^{2}k}{\partial {\omega}^{2}}$ is negligible (which will be justified later). With the undepleted pump approximation, Eq. (9) can be integrated to get the power spectrum of the THz field

Here we assumed *χ*
^{(2)} = 2*d*
_{eff}, where *d*
_{eff} = (2/*π*)*d*
_{OR} is an effective QPM nonlinear coefficient; *d*
_{OR} corresponds to the optical rectification process *d*
_{OR}(0=ω-ω) and is derived from electro-optic coefficient *r*_{ijk}
using the relation [12] *d*
_{jkl}= - *n*
^{4}/4*r*
_{jlk}, where *n* is the optical refractive index. In GaAs, for example, *r*
_{14}=1.5pm/V [14], corresponding to the nonlinear coefficient *d*
_{OR} = 47 pm/V.

If the nonlinear crystal is long enough, terahertz radiation will be emitted in the form of narrow-band spectrum centered at Ω_{0}, corresponding to Δ*k*=0 condition

where Δ*n* = *n*_{THz}
- ${n}_{\mathit{\text{opt}}}^{\mathit{\text{gr}}}$
is the index mismatch. By differentiating (11) we get $\frac{d\left(\Delta k\right)}{d\Omega}=\frac{\Delta n}{c}$ and the phase-matching acceptance bandwidth based on the condition Δ*kL*/2 = *π*

We assumed that *n*_{THz}
is nearly constant, which is true for frequencies well below the lowest phonon resonance. For GaAs, this resonance is at 8.1 THz [15], *n*_{THz}
≈3.6 [16], ${n}_{\mathit{\text{opt}}}^{\mathit{\text{gr}}}$
=3.41 (for the 2.1-μm optical pump) [17], *Δ*n=0.19, and acceptance bandwidth for *L*=1cm crystal is Δ${\nu}_{\mathit{\text{THz}}}^{\mathit{\text{accept}}}$
=53 cm^{-1}. In the absence of quasi phase-matching, interaction between the optical and THz waves is limited to the coherence length

corresponding to the Δ*kl*_{c}
= *π* condition.

It is worth mentioning that *backward emission* of THz wave is also possible. In this case, the phase-matching condition becomes

Optical-to-THz energy fluence efficiency for plane waves (PW) is

The pump fluence is given by

where *n*
_{2} is the optical refractive index. The THz fluence is

where we used Parseval’s theorem: $\underset{-\infty}{\overset{\infty}{\int}}{\mid f\left(t\right)\mid}^{2}\mathit{dt}=2\pi \underset{-\infty}{\overset{\infty}{\int}}{\mid f\left(\Omega \right)\mid}^{2}d\Omega $.

From Eqs. (12) and (18–19) we get

where we replaced Δ*kL*/2 with Δ*n*(Ω - Ω_{0})*L*/2*c*. For *L* ≫ *l*
_{c}, which is equivalent to the condition ΔΩ^{accept} ≪ Ω_{0}, sinc^{2} function under the integral dominates and we obtain

where Ω_{0} is given by Eq. (13). The reduction factor

reflects the fact that the optical pulse should be short enough, so that its spectrum span is larger than the THz frequency Ω_{0}. For very short optical pulses, *τ*Ω_{0} < 1, *g*
_{1}≈1 and optical-to-THz conversion efficiency depends *on pulse fluence only*, not intensity. Figure 2 shows the reduction factor *g*
_{1} as a function of the product ν_{Thz}τ. For example at ν_{THz}τ=0.1, *g*
_{1}= 0.91, close to unity, and experiences little change as the pulse duration is further decreased.

## 3. Plane wave analysis, picosecond pulses

Let us now regard as a pump, bandwidth-limited pulses with longer (pico- or nanosecond) duration τ, such that ν_{THz}τ>1. In this case, the spectrum of a single pulse is narrow and to generate THz output, two different pump pulses need to be mixed to achieve difference frequency generation (DFG). Assume that two gaussian bandwidth-limited optical pulses (plane waves) at frequencies ω_{2} and ω_{3} with equal pulse widths propagate collinearly and generate THz wave centered at Ω_{0}=ω_{3} - ω_{2}. Assume that the electric fields (*i*=2,3) are in the form

then, similar to Eq. (4), Fourier transform is given (ω>0) by

The Fourier component of nonlinear polarization at THz frequency Ω is

Integrating Eq. (6) in the limit of no absorption and no pump depletion, we get the power spectrum of the THz field

where Δ*k* is given by (11).

Suppose that the QPM orientation-reversal period Ό is such that the sinc^{2} peak is centered exactly at Ω_{0}=ω_{3} - ω_{2}. In this case Δ*kL*/2 can be replaced by (Ω-Ω_{0})Δ*nL*/2*c*. Using Parseval’s theorem, we find the optical-to-THz energy fluence conversion efficiency with respect to one of the two pump pulses (at ω_{2})

where Ω’=Ω-Ω_{0}

The temporal walk-off length between the optical and THz pulses can be introduced as

which is proportional to the length in a crystal at which the optical pulse walks away in time, with respect to THz wave, by its width τ, due to difference in propagation velocities. It is analogous to the Boyd-Kleinmann’s [18] birefringent aperture length in the theory of second harmonic generation *l*_{a}
= √*πw*
_{0}/*ρ* (*w*
_{0} is beam radius, and ρ is the birefringent walk-off angle). Now (27) can be expressed as

In the limit of long pulses, *l*
_{w}≫*L*, exponential function under the integral dominates and (29) becomes

where *n*
_{2} and *n*
_{3} are refractive indices at ω_{2} and ω_{3} and *I*
_{3} is the peak pump intensity. This formula is similar to the well-know expression [19] for the CW difference frequency generation, with *I*
_{3}/√2 playing the role of time-averaged pump intensity.

In the limit of short pulses, *l*
_{w}≪*L*, (29) becomes

Thus, for *l*
_{w}≪*L*, the *L*
^{2} term is replaced by *Ll*
_{w}, in full analogy with the case of second harmonic generation with the spatial walk-off [18], where *L*
^{2} is replaced by *I*
_{a} for *l*
_{a}≪*L*. Also, conversion efficiency can be rewritten in terms of coherence length *l*
_{c}=π*c*/ΩΔ*n* and energy fluence *F*
_{3}

This shows that at *l*
_{w}≪*L*, terahertz conversion efficiency is a function of fluence only. Besides, it is equal to the conversion efficiency for the case of a femtosecond pulses with ν_{THz}τ≪1. The only difference is that in a two-color picosecond case, in order to get the same energy per THz pulse, one needs to have twice total energy (*U*
_{0} in each of the beams), as compared to *U*
_{0} in the femtosecond case.

In the case of intermediate pulse durations, we can use (32) with a reduction factor *g*
_{2}(*l*
_{w}/*L*),

where

Figure 3 is the plot of the reduction factor *g*
_{2} as a function of *l*
_{w}/*L*. In many cases, it is desirable to have longer pulses to suppress high-order nonlinear optical effects, even at the expense of some loss in efficiency. Thus setting *l*
_{w}/*L*=1 (*g*
_{2}=0.69) might be a good compromise between efficiency and pump intensity. For an *L* = 1cm GaAs and pump at 2.1 μm, the *l*
_{w}/*L* =1 condition corresponds to the pulse duration of 3.6 ps. At longer pulses, the THz efficiency will decline; however it will not be improved dramatically if the pulses are made shorter.

## 4. Chirped stretched femtosecond pulses

Narrow-band THz radiation can also be generated by mixing two linearly chirped optical pulses (Fig. 4). Overlap of two relatively delayed optical pulses produces a constant beat frequency, which is proportional to the time delay and the amount of frequency chirp. Weling *et al*. [20] demonstrated this with photoconducting antennas and 23-ps-long pulses and showed that the generated THz frequency exhibited a linear dependence on the time delay. The same technique can be applied to THz generation using optical rectification.

Assume that the two pulses are of equal intensities. The electric field of the first optical pulse is in the form

and that of the second (delayed) pulse is

Then the nonlinear polarization at the THz beat frequency is

$$\phantom{\rule{9.5em}{0ex}}=\mathrm{Re}\left\{{\epsilon}_{0}{\chi}^{\left(2\right)}{E}_{0}^{2}\mathrm{exp}\left[-\frac{2}{{\tau}^{2}}\left({t}^{2}+\Delta \mathit{tt}+\frac{\Delta {t}^{2}}{2}\right)+i\left(2b\Delta \mathit{tt}+b\Delta {t}^{2}\right)\right]\right\}$$

Its Fourier transform (Ω>0)

$$={\epsilon}_{0}{\chi}^{\left(2\right)}\frac{{E}_{0}^{2}\tau}{2\sqrt{2\pi}}\mathrm{exp}(-\frac{{\tau}^{2}{\mathrm{\Omega \text{'}}}^{2}}{8}-\frac{1}{2}\frac{\Delta {t}^{2}}{{\tau}^{2}})\mathrm{exp}\left(i{\phi}_{1}\right)\phantom{\rule{10em}{0ex}}$$

where Ω’=2*b*Δ*t* - Ω and φ_{1} = -Δ*t*Ω’2+*b*Δ*t*
^{2}.

Similar to Eq. (29), we get THz fluence conversion efficiency with respect to one of the two pump pulses

Exponential function here peaks at Ω’=0, that is when Ω=2*b*Δ*t*. If we assume that the sinc_{2} function peaks at the same frequency (Ω”=0), then Eq. (39) becomes

Here *g*
_{2} is the reduction factor given by Eq. (34); the exponential term can be regarded as another reduction factor, associated with the temporal overlap of the two pulses, one of which is delayed by Δ*t*. Suppose that the two chirped pulses with the pulsewidth τ=τ_{2} are created by stretching in time a much shorter (femtosecond) pulse with the pulsewidth τ_{1}. It is easy to show [21] that in this case *b*=1/τ_{1}τ_{2} and exp(-(Δ*t*/τ_{2})^{2})= exp(-(Ωτ_{1}/2)^{2}). Interestingly, this reduction factor is the same as *g*
_{1}(τ_{1}) in (22); thus (40) becomes

We see that in the case of linearly chirped optical pulses, THz conversion efficiency is, again, a function of fluence only, provided that pulse durations τ_{1} and τ_{2} are short enough.

## 5. Optimal length of the EO crystal

From the simple plane-wave analysis above, we see that the optical-to-THz conversion efficiency, in the optimized case, is proportional to *L*. If we take into account THz absorption in the crystal (which is usually much larger than the optical absorption) but still neglect pump depletion, we get

where α is the THz intensity absorption coefficient and *L*
_{eff} is an effective length. When *L* → ∞, *L*
_{eff} → 1/α. Choosing *L*=1/α will give us *L*
_{eff}=0.63/α. In general, one can introduce another reduction factor, associated with the absorption and write

For *L*=1/α, *g*
_{3}=0.63.

## 6. Optimal focusing

To maximize the THz efficiency, one needs to focus pump beams. In the near field approximation, when the focusing is loose and diffraction can be neglected, the optical-to-THz conversion efficiency (OR, femtosecond pulses) with respect to the pump pulse energy *U*
_{pu} can be obtained by integrating Eq. (21) over the transverse coordinate *r*(*E*_{pu}
(*r*) ~ exp(-*r*
^{2}/*w*
^{2}))

where *w* is the gaussian pump beam size. Following Ch.5 of [18], which considers the DFG case, we can characterize the focusing strength by a focusing parameter *ξ* = *L*/*k*
_{1}
*w*
^{2} = *λ*
_{1}
*L*/2*πn*
_{1}
*w*
^{2}, where *λ*
_{1} is the THz wavelength and *n*
_{1} is the THz refractive index (Boyd-Kleinman’s theory itself [18] is not applicable here since THz field is not a resonant field and its distribution is not defined a priori by an optical cavity). Morris and Shen [22] developed a theory of far-infrared generation by optical mixing of focused laser beams, based on Fourier analysis with respect to transverse *k*-vector components, and have found that focusing of the pump beams appreciably enhances the far-infrared output despite the strong far-infrared diffraction. For example in a 1-cm-long GaAs crystal and an output wavelength 100μm, the optimal focal-spot size (for the optimized phase-matching condition) was found to be around *w*=20 μm (*ξ*=110), that is less than the THz wavelength.

THz conversion efficiency for OR with femtosecond pulses and focused beams can be expressed in the form

and, similarly, for DFG with picosecond pulses, by

Here *h*(*ξ*), by analogy with [18], is an enhancement factor due to pump focusing, such that *h*(ξ)=ξ, at ξ≪1 (plane waves). Enhancement factor *h*(ξ) obtained from the calculations of Morris and Shen [22] is shown on Fig. 5, along with the enhancement factor for plane-waves. Fig. 5 (dots) also shows results of our calculations of the enhancement factor, based on the Green’s function method [23,24], where the far field solution was obtained by integration over the nonlinear polarization distribution in real space; these results are in excellent agreement with the Morris-Shen curve.

There are however limitations on the very tight focusing. (i) When pump beam waist *w* is too small, THz output extends over a large span of angles θ to the normal of the crystal; when θ exceeds the total internal reflection angle θ_{max} of the material, THz transmission falls to zero. This leads to the condition √2*λ*/*πn*
_{1}
*w* < 2θ_{max}. For GaAs, for example, θ_{max} = 16°, and for 1.5 THz frequency, this corresponds to *w* > 45 μm. (ii) In the case of tight focusing, different Fourier components of the transverse *k*- vector have different phase-matching conditions and it appears [22] that the THz output is maximized when the far field distribution is in the form of a hollow cone. The phase-matching condition for each angle θ is

where Δ*k*
_{0} is a collinear wave-vector mismatch. For the high beam quality (solid vs hollow cone) we can require that variation of Δ*k*
_{z} due to variation of θ is small, so that δ(Δ*k*
_{z})*L*<π (corresponding to half of the QPM acceptance bandwidth). This leads to the condition *k*
_{1}${\mathrm{\theta}}_{d}^{2}$
*L*/2 ≤ π where θ_{d} is given by the diffraction: θ_{d} = √2λ_{1}/2*πnw*. Thus we get

The far field intensity profiles of THz radiation at different ξ(GaAs, *L*=1cm) based on our calculations using the Green’s function method is shown in the inset to Fig. 5 and is consistent with Eq. (48); ξ≈1 can be regarded as a good compromise between conversion efficiency and the beam quality.

From Eqs. (45–46) one can see that the THz conversion efficiency does not depend on the crystal length *L*. The length affects however the spectral width of the THz pulse; also, it is more advantageous to have longer crystals (on the order of 1/α), since in this case the beam waist will be larger and the peak intensity smaller. This is important from the viewpoint of reducing unwanted higher order nonlinear effects (see below).

As a numerical example, regard a 100-fs-long optical pump pulse at 2.1 μm, 1.5 THz output frequency, GaAs crystal length *L*=1/α =9.5 mm (α=1.05 cm^{-1}), and pump beam waist *w*=290 μm (ξ=1). In this case reduction factors are *g*
_{1}=0.8, *g*
_{3}=0.63, *h*(ξ)=0.82 and from (45) we obtain η_{THz}= 3.95×10^{-4}/ μJ. For the 2.5 THz output (α=4.3 cm^{-1}), for *L* =2.3 mm and *w*=110 μm (ξ=1), *g*
_{1}=0.54, *g*
_{3}=0.63, *h*(ξ)=0.82, and η_{THz}=1.23×10^{-3} / μJ.

## 7. Cascading and red shift

THz wave generation via OR is a parametric process of self-mixing in which a photon from the blue (high-frequency) wing of the femtosecond optical pulse decays into a THz photon plus a red-shifted photon, corresponding to the low-frequency wing of the same optical pulse. From the photon energy conservation argument, it follows (if we neglect losses) that the center of weight of the optical pulse spectrum will be red-shifted by Δω/ω_{0} ~ *η*_{THz}
, where ω_{0} is the central optical frequency. When optical-to-THz photon conversion efficiency approaches 100%, the red shift will be on the order of THz frequency Ω. Once the optical pulse becomes red-shifted, it can still contribute to THz generation ‒ the same process of cascaded optical down-conversion continues to transfer optical energy to lower frequencies, as long as the phase mismatch Δ*k* in Eq. (11) is small. Accordingly, cascaded down-conversion will be the most efficient when the pump wavelength is close to the point of zero group velocity dispersion. Quantitatively, the number of cascading cycles can be expressed as *N*= ½ (acceptance bandwidth) / (teraherz frequency). Here acceptance bandwidth is with respect to the pump frequency and can be found by differentiating (11): *d*(Δ*k*)/*dω*_{pump}
= (Ω/*c*)*dn*_{gr}
/*dω*, from which we get

From Fig. 6 we can see that the number of THz cascading cycles in GaAs at 2-3.5 μm pump can be > 10. At pump wavelength near 6.6 μm, where the group velocity dispersion reaches zero, N can be even higher. Thus, THz conversion efficiency can be significantly above the Manley-Rowe limit. The possibility of overcoming quantum-defect-related limitations on the efficiency of THz wave difference frequency generation, for the case of mixing two discrete near infrared frequencies, was suggested by Cronin-Golomb [25].

The red shift of the optical pump pulse can also be understood from a different standpoint. Gustafson *et al*. [26] have pointed out that rectified field which is generated during the propagation of optical pulses through electro-optic crystals can produce considerable phase modulation of the optical pulse itself, especially when the rectified pulse travels at the same speed as the optical pulse within the crystal. This cascaded Kerr-like nonlinearity occurs because of back action of the optical rectified field upon the pump wave through the electro-optical effect [27,28].

Figure 7 illustrates the process of red shift for the case of single-cycle terahertz pulse generation. Through the EO effect, rectified field (Fig. 7(b)) produces a change in the refractive index of the medium δ*n* ~ *r*
_{ij}
*E*
_{THz}(*t*,*z*) which causes the time-dependent phase shift δϕ(*t*,*z*) (Fig. 7(c)). The amplitude of this phase shift is proportional to the integral of the rectified field over the crystal length; since the rectified field grows linearly with the length,
the phase shift is proportional to the crystal length squared [26]. The frequency shift δω=-∂δϕ(*t*,*z*)/∂*t* (Fig. 7(d)) is also proportional to the crystal length squared. The central portion of the optical pulse experiences red shift, independent on the sign of EO coefficient. This ‘microscopic’ consideration of the red shift gives the same dependence as before: it scales as THz conversion efficiency (proportional to the EO coefficient squared and crystal length squared) for the index-matched single-cycle terahertz pulse generation. In the case of multicycle THz generation with periodically inverted crystals, the optical pulse will see in average the same induced phase shift as shown in Fig. 7(b), since the relative phase between the optical and THz pulse is corrected at regular intervals via a structural periodicity of the nonlinear medium.

## 8. The role of higher order nonlinear effects

Two-photon absorption (2PA), three-photon absorption (3PA), and nonlinear refraction index (NRI), n_{2}, can severely limit the maximum THz conversion efficiency in the OR or DFG process: in the case of 2PA and 3PA, the pump beam will create free carriers which strongly absorb THz radiation; NRI, on the other hand, affects the pump pulse itself via self-phase modulation and self-focusing. As a necessary condition for achieving cascading effects in THz generation, one needs to be close to 100% in a single-stage optical-to-THz photon efficiency, which requires for GaAs pump intensities on the order of 100-1000 GW/cm^{2} (fs pulses) and 1–100 GW/cm^{2} (ps pulses). It is very advantageous to use longer pump wavelengths to avoid 2PA effects which are usually strong in semiconductors (e.g. for GaAs, β=26 cm/GW at 1.06 μm [29]). This dictates that pump wavelength should be above 1.74 μm (GaAs) in which case the dominant nonlinearities are 3PA and NRI (theoretically predicted values for GaAs are correspondingly γ≈0.2 cm^{3}/GW2 [30] and n_{2}≈2·10^{-4} cm^{2}/GW [31]). Also, it is more likely that the highest THz conversion efficiencies (> 100% by photons, with cascading) will be achieved in the picosecond, rather than femtosecond regime: since THz conversion efficiency is fluence-sensitive, peak intensity in the former case will be much smaller, and 3PA and NRI effects less pronounced.

## 9. Conclusion

We show that optically pumped periodically-inverted EO crystals have potential for generating narrow-bandwidth THz wave packets with high conversion efficiency. For sufficiently short (femtosecond) single-beam optical pulses with τΩ<1 (optical rectification), THz conversion efficiency scales with the pulse energy and does not depend on pulse duration. In the case of mixing two picosecond pulses (bandwidth-limited or chirped), with pulse durations such that the temporal walk-off length is smaller than the length of the crystal. again, conversion efficiency is pulse-energy dependent, and has the same scaling factor as in the case of femtosecond pulses. Also, cascaded THz frequency down conversion is possible, allowing achieving > 100% optical-to-THz photon conversion efficiency.

## Acknowledgments

We would like to thank Martin Fejer, Gennady Imeshev, Vladimir Kozlov and Yun-Shik Lee for helpful discussions. Also, special thanks to Y.-S. Lee for the animated Fig. 1. This work was sponsored by DARPA under AFOSR Grant FA9550-04-1-0465.

## References and links

**1. **T. Yajima and N. Takeuchi, “Far-infrared difference-frequency generation by picosecond laser pulses,“ Jpn. J. Appl. Phys. **9**, 1361–1371 (1970). [CrossRef]

**2. **K.H. Yang, P.L. Richards, and Y.R. Shen, “Generation of far-infrared radiation by picosecond light pulses in LiNbO_{3},” Appl. Phys. Lett. **19**, 320–323 (1971). [CrossRef]

**3. **L. Xu, X.-C. Zhang, and D. H. Auston, “Terahertz beam generation by femtosecond optical pulses in electro-optic materials,” Appl. Phys. Lett. **61**, 1784–6 (1992) [CrossRef]

**4. **B. Ferguson and X.-C. Zhang, “Materials for terahertz science and technology,” Nature Materials **1**, 26–33 (2002). [CrossRef]

**5. **A. Bonvalet, M. Joffre, J.-L. Martin, and A. Migus, “Generation of ultrabroadband femtosecond pulses in the mid-infrared by optical rectification of 15 fs light pulses at 100 MHz repetition rate,” Appl. Phys. Lett. **67**, 2907–2909 (1995). [CrossRef]

**6. **R. A. Kaindl, F. Eickemeyer, M. Woerner, and T. Elsaesser, “Broadband phasematched difference frequency mixing of femtoseconds pulses in GaSe: Experiment and theory,” Appl. Phys. Lett. **75**, 1060–1062 (1999). [CrossRef]

**7. **Peter H. Siegel, “Terahertz Technology,“ IEEE Transactions on Microwave Theory and Techniques **50**, 910–28 (2002). [CrossRef]

**8. **T. J. Carrig, G. Rodriguez, T. S. Clement, and A. J. Taylor, “Scaling of terahertz radiation via optical rectification in electro-optic crystals,” Appl. Phys. Lett. **66**, 121–3 (1995). [CrossRef]

**9. **A. G. Stepanov, J. Kuhl, I. Z. Kozma, E. Riedle, G. Almási, and J. Hebling, “Scaling up the energy of THz
pulses created by optical rectification,” Optics Express **13**, 5762–68 (2005). [CrossRef] [PubMed]

**10. **Y.-S. Lee, T. Meade, V. Perlin, H. Winful, T. B. Norris, and A. Galvanauskas, “Generation of narrow-band terahertz radiation via optical rectification of femtosecond pulses in periodically poled lithium niobate,” Appl. Phys. Lett. **76**, 2505–7 (2000). [CrossRef]

**11. **K. L. Vodopyanov, M. M. Fejer, D. M. Simanovskii, V.G. Kozlov, and Y.-S. Lee, “Terahertz-wave generation in periodically-inverted GaAs,” Conference on Lasers and Electro Optics, May 2005, Baltimore MD, Technical Digest (Optical Society of America, Washington DC, 2005), paper CWM1.

**12. **A. Yariv, *Quantum Electronics*, (Wiley, New York, 3rd edition, 1988), Chapter 16.

**13. **A. Nahata, A. S. Weling, and T. F. Heinz, “A wideband coherent terahertz spectroscopy system using optical rectification and electro-optic sampling,” Appl. Phys. Lett. **69**, 2321–23 (1996). [CrossRef]

**14. **V. G. Dmitriev, G. G. Gurzadyan, and D. N. Nikogosyan, *Handbook of Nonlinear Optical Crystals*, (Springer, Berlin, 1997).

**15. **W.J. Moore and R.T. Holm, “Infrared dielectric constant of gallium arsenide,” J. Appl. Phys. **80**, 6939–42 (1996). [CrossRef]

**16. **D. Grischkovsky, S. Keiding, M. van Exter, and Ch. Fattinger, “Far-infrared time-domain spectroscopy with terahertz beams of dielectrics and semiconductors,” J. Opt. Soc. Am. B **7**, 2006–2015 (1990). [CrossRef]

**17. **T. Skauli, P. S. Kuo, K. L. Vodopyanov, T. J. Pinguet, O. Levi, L. A. Eyres, J. S. Harris, and M. M. Fejer, “Determination of GaAs refractive index and its temperature dependence, with application to quasi-phasematched nonlinear optics,” J. of Appl. Phys. **94**, 6447–55 (2003). [CrossRef]

**18. **G.D. Boyd and D.A. Kleinman, “Parametric interaction of focussed gaussian light beams,” J. Appl. Phys. **39**, 3597–3639 (1968). [CrossRef]

**19. **R.L. Byer and R.L. Herbst, “Parametric oscillation and mixing,” in Topics in Applied Physics: *Nonlinear Infrared Generation*, ed. by Y.R. Shen (Springer, Berlin, 1977), vol. 16, p. 81–137. [CrossRef]

**20. **A. S. Weling, B. B. Hu, N.M. Froberg, and D.H. Auston, “Generation of tunable narrow-band THz radiation from large aperture photoconducting antennas,” Appl. Phys. Lett. **64**, 137 (1994). [CrossRef]

**21. **A. E. Siegman, *Lasers*, (University Science Books, Mill Valley, 1986), Ch.9.

**22. **J.R. Morris and Y.R. Shen, “Theory of far infrared generation by optical mixing,” Phys. Rev. A **15**, 1143–56 (1977). [CrossRef]

**23. **Y.R. Shen, “Far-infrared generation by optical mixing,” Prog. Quant. Electr. **4**, 207–232 (1976) [CrossRef]

**24. **S. Guha, “Focusing dependence of the efficiency of a singly resonant optical parametric oscillator,” Appl. Phys. B **66**, 663–675 (1998). [CrossRef]

**25. **M. Cronin-Golomb, “Cascaded nonlinear difference-frequency generation of enhanced terahertz wave production,” Opt. Lett. **29**, 2046–48 (2004). [CrossRef] [PubMed]

**26. **T.K. Gustafson, J.-P.E. Taran, P.L. Kelley, and R.Y. Chiao, “Self-modulation of picosecond pulses in electro-optic crystals,” Opt. Commun. **2**, 17–21 (1970). [CrossRef]

**27. **J.-P. Caumes, L. Videau, C. Touyez, and E. Freysz, “Kerr-like nonlinearity induced via terahertz generation and the electro-optic effect in zinc blende crystals,” Phys. Rev. Letters **89**, 047401 (2002). [CrossRef]

**28. **Y.J. Ding, “Efficient generation of high-power quasi-single-cycle terahertz pulses from a single infrared beam in a second-order nonlinear medium,” Opt. Lett. **29**, 2650–52 (2004). [CrossRef] [PubMed]

**29. **A.A. Said, M. Sheik-Bahae, D.J. Hagan, T.H. Wei, J. Wang, J. Young, and E.W. Van Stryland, “Determination of bound-electronic and free-carrier nonlinearities in ZnSe, GaAs, CdTe, and ZnTe,” J. Opt. Soc. Am. B **9**, 405–414 (1992). [CrossRef]

**30. **B.S. Wherrett, “Scaling rules for multiphoton interband absorption in semiconductors,” J. Opt. Soc. Am. B **1**, 67–72 (1984). [CrossRef]

**31. **M. Sheik-Bahae, D.C. Hutchings, D.J. Hagan, and E.W. Van Stryland, “Dispersion of bound electronic nonlinear refraction in solids,” IEEE J. of Quant. Electron. **27**, 1296–1309 (1991). [CrossRef]